A Survey on Personality-Aware Recommendation Systems
Sahraoui Dhelim, Nyothiri Aung, Mohammed Amine Bouras, Huansheng Ning, Erik Cambria
SSUBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 1
A Survey on Personality-Aware RecommendationSystems
Sahraoui Dhelim, Nyothiri Aung, Mohammed Amine Bouras, Huansheng Ning and Erik Cambria.
Abstract —With the emergence of personality computing as anew research field related to artificial intelligence and personalitypsychology, we have witnessed an unprecedented proliferation ofpersonality-aware recommendation systems. Unlike conventionalrecommendation systems, these new systems solve traditionalproblems such as the cold start and data sparsity problems.This survey aims to study and systematically classify personality-aware recommendation systems. To the best of our knowledge,this survey is the first that focuses on personality-aware recom-mendation systems. We explore the different design choices ofpersonality-aware recommendation systems, by comparing theirpersonality modeling methods, as well as their recommendationtechniques. Furthermore, we present the commonly used datasetsand point out some of the challenges of personality-awarerecommendation systems.
Index Terms —Recommendation system, Personality comput-ing, personality traits, Big-five model, personality-aware, socialcomputing, collaborative filtering.
I. I
NTRODUCTION P ERSONALITY Computing is the interdisciplinary studyfield that focuses on the integration of personality psy-chology theories with computing systems. It has been proventhat leveraging personality theories could help to tackle someof the well-known problems in computer science. Personalitycomputing has been applied in many domains and researchdirections, and the number of scientific publications withinthe scope of personality computing has dramatically increasedwithin the last decade. The integration of user personalitytraits into the computing system has created new researchdirections, such as automatic personality recognition, andhelped to accelerate existing research directions as well, suchas recommendation systems, and human-robot interaction re-search. Personality computing has enabled recommendationsystems to understand the users’ preferences from a differ-ent perspective. A new type of recommendation system thatleverages the user’s personality trait to improve the recom-mendations had emerged. This group of systems is known asPersonality-aware recommendation systems. This new type ofrecommendation systems has proven effective in solving theproblem of conventional recommendation systems. Such as thecold-start problem, when the system does not have much dataabout the preferences of the user, free-riders problem and thedata sparsity problem, to name a few.
Sahraoui Dhelim Nyothiri Aung Mohammed Amine Bouras HuanshengNing are with School of Computer and Communication Engineering, Univer-sity of Science and Technology Beijing, 100083, Beijing, China.Erik Cambria is with Nanyang Technological University, SingaporeCorresponding author: Huansheng Ning ([email protected]).
In the recent few years, we have witnessed a rapid prolif-eration of personality-aware recommendation systems. Whileall of these recommendation systems incorporate the user’spersonality traits in the recommendation process, however,these systems use different recommendation techniques, andthey are designed for different recommended content. There-fore, in this paper, we conduct a comprehensive review ofthe literature of personality-aware recommendation systems.Few works surveyed some research direction in the field ofpersonality computing. In 2014, Vinciarelli and Mohammadi[1] surveyed the publications that used the user’s personalityin computing systems, and they coined the term PersonalityComputing. In 2017, Kaushal and Patwardhan [2] surveyedthe literature on automatic personality recognition from onlinesocial networks. Similarly, in 2019, Mehta et al. . [3] surveyedthe literature on deep-learning-based personality automaticpersonality recognition. However, as far as we know, weare the first who survey the literature of personality-awarerecommendation systems. In Tables I, we list some of therecent surveys in the field of personality computing, alongwith their focus scope and publication year.The remainder of this paper is organized as follows:In Section 2, we show the main differences betweenconventional recommendation systems and personality-awarerecommendation systems by explaining the main componentof the latter. While in Section 3, we systematically classifythe existing personality-aware recommendation system basedon the used recommendation technique. In Section 4, wereview some of the works that proposed personality-awarerecommendation systems in the last few years. Whereas inSection 5, we present some of the commonly used datasetsand benchmarks related to personality-aware recommendationsystems. In Section 6, we discuss some of the challenges thatface personality-aware recommendation systems and also listsome of the open issues and research challenges. Finally, weconclude this survey in Section 7.II. P
ERSONALITY - AWARE RECOMMENDATION SYSTEMS
Historically, recommendation systems are divided into threemain categories, collaborative filtering approaches, contentfiltering approaches and hybrid filtering approaches. Collabo-rative filtering is inspired by the fact that “people who agree onthe past, probably will agree in the future”. In practice, in orderto recommend new items to a given user u x , collaborativefiltering systems determine a group of users that have a similarrating with user u x , these users are called the neighbors of user u x . After finding the group of neighbors, the system finds the a r X i v : . [ c s . I R ] J a n UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 2
TABLE I: Related surveys and reviews
Research field Publication Scope description Year Note
Personalitycomputing Vinciarelli and Mo-hammadi [1] A general survey on personality computing 2014 Vinciarelli and Mohammadi coined the term“Personality Computing”Wright [4] Commentary about [1] 2014 Wright explain his perspective about person-ality computing as a personality psychologistVinciarelli and Mo-hammadi [5] A complement of [1] by the same authors. 2014 Vinciarelli and Mohammadi replied on thecommentary of Wright [4], and discussed thefuture direction of Personality Computing.Automaticpersonalityrecognition Jacques et al. [6] A survey on vision-based personality detec-tion 2019 The authors have surveyed only image-basedand video-based personality detectionKaushal et al. [2] A survey on user personality detection fromonline social networks 2018Mehta et al. [3] A survey on deep learning based personalitydetection. 2019Bhavya et al. [7] A review on deep learning based personalitydetection from online social networks. 2019 These works surveyed textual, as well asnon-textual (image, video, voice) personalitydetection.Kedar et al. [8] A review on various approaches used forpersonality assessment 2015Finnerty et al. [9] A review on the data sources and the featuresused for automatically infer personality 2016Azucar et al. [10] Review on Big 5 personality traits fromdigital footprints on social media 2017Dandannavara et al. [11] A review on text-based personality predic-tion from online social networks 2019 The authors have reviewed only text-basedpersonality detectionPersonality inhuman-robotinteraction Santamaria et al. [12] A review on research methods for measuringpersonality in human-robot interaction 2017Robert [13] A review on personality in human-robot in-teraction 2019 These works focus on robots’ artificial per-sonality traits and the application of person-ality computing in human-robot interactionRobert et al. [14] A review on personality in human-robot in-teraction 2020Mou et al. [15] A review on the personality of robots 2020 set of items that share a high rating among these neighbors,and subsequently recommend these items to user u x . Whilecontent filtering approaches, compute the similarity betweenprevious matched items and the suggested items, regardlessof the neighbors’ ratings. Finally, hybrid approaches use acombination of these two techniques. Similar to the conven-tional recommendation systems, personality-aware recommen-dation systems also use similar recommendation techniques,the only difference is that they add the user’s personalityinformation in the recommendation process. In Figure 1 andFigure 2, we show the main differences between conventionaland personality-aware recommendation systems. Conventionalrecommendation systems mainly have three stages. Firstly,the rating phase, where the user expresses her interests byrating some items. The second stage is the filtering phase,either collaborative filtering, content filter or hybrid filteringas mentioned above. Finally, at the recommendation phase, thesystem recommends the items yielded by the filtering phase.As shown in Figure 2, personality-aware recommendationsystems add two more phases before the rating phase andchange the filtering stage as well. Personality measurementphase, the system assesses the personality type of the usersusing a personality assessment questionnaire that the usersneed to answer during registration, or by applying an auto-matic personality recognition scheme on the users’ previouslyavailable data, such as online social network data. While in thepersonality matching phase, the system tries to match the userpersonality type with relevant items, this could be done eitherusing lexical matching by linking the textual description of theitems with the associated personality types, or using a fine- Fig. 1: Conventional recommendation systemsFig. 2: Personality-Aware recommendation systems UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 3 grained rules that can match items with personality types. Itis worth noting that at personality matching phase, the systemdoes not have any information about the users’ ratings, whichhelp to alleviate the effects of cold-start problem, one of themost challenging problems in the literature of recommenda-tion systems. Personality-aware recommendation systems alsochange the filtering phase, by incorporating the personalityinformation in the similarity measurement to determine theneighbors of each user.
A. Personality measurement
The personality measurement is the most important phase ofpersonality-aware recommendation system, as any misidenti-fication of the user personality type could negatively influencethe accuracy of the recommendation system. In personalitycomputing, there are two main methods for personality mea-surement, personality assessment questionnaires and automaticpersonality recognition (APR). Generally, questionnaires basedpersonality measurement is more accurate than APR. However,in the context of personality computing, APR are relativelyeasier to conduct, as they can be applied to the user’s existingdata, without the need to burden the user with long question-naires. In this subsection, we discuss the different personalityassessment questionnaires and APR techniques and classifysome of the existing personality-aware recommendation sys-tems based on the their personality assessment method.
1) Personality assessment questionnaires:
In the study fieldof personality psychology, self-report personality question-naires have been widely used to reveal personality differencesamong individuals. Questionnaires in which people estimatetheir character and behaviors are the most commonly usedmean for personality assessment. The answers are typicallyin the format of five-level Likert scale (strongly agree, agree,neither agree or disagree, disagree and strongly disagree).There are many personality questionnaires of variouslengths (number of items). The widely used long question-naires include NEO Five-Factor Inventory (NEO-FFI, 60items) [16], NEO-Personality-Inventory Revised (NEO-PI-R,240 items) [17], International Personality Item Pool (IPIP-NEO, 300 items) and the Big-Five Inventory (BFI, 44 items)[18]. In the context of personality-aware recommendationsystem, these questionnaires could be directly solicited fromthe users during registration. Filling a long questionnaire is atime-consuming task, the user might get bored and may not fillthe questionnaire carefully, which could lead to an irrelevantrecommendation in the future. Therefore, short questionnaires[18], [19], [20] (5-10 items) are preferred in personality-awarerecommendation systems, as these questionnaires are mucheasier to fill. The most prominent short questionnaires areBFI-10 (a short version of BFI with 10 items) and Ten-ItemPersonality Inventory (TIPI, 10 items) [19]. Table II shows theitems of BFI-10.There are two main drawbacks of self-assessment question-naires. The first limitation is the self-bias problem [21], whenthe subjects tend to give the wrong answer to some of theundesired social characteristics in certain circumstances. Forexample, when answering a recruitment personality question- TABLE II: The BFI-10 Personaltiy questionnaire
Item Question Dimension
TABLE III: Questionnaire types in personality-awarerecommendation systems
Recommendationsystem Personalityassessmentquestionnaire Itemcount [23], [24], [25],[26], [27], [28],[29], [30], [31],[32], [33], [34],[35] TIPI 10[36], [37], [38],[39], [40] IPIP 50-version 50[41], [42], [43],[44], [45], [46] BFI 44[47], [48] NEO-FFI 60[49], [50], [51], [52] Big-five marker scale(BFMS) [53] 100[54], [55], [56] BFI-10 10[57] IPIP-NEO-60 60[58] IPIP 44-version 44[59] IPIP 336-version 336[60] IPIP 100-version 100[61] FIPI 5 naire, most of the subjects gives inaccurate answers to ques-tions like “I get nervous easily”, “I tend to be lazy”, becausethese are undesired characters in employees. The self-biasdoes not affect personality-aware recommendation systems,because users have no benefit in misleading the system. Thesecond drawback is known as the reference-group effect [22],in which the answers given by the subject is relative to hissurrounding environment. For example, an introvert engineermight think he is extravert if he is surrounded by a group ofeven more introvert engineer friends. In Table III we list thepersonality assessment questionnaires that were used in therecent personality-aware recommendation system.
2) Automatic personality recognition:
The assessment ofusers’ personality using a questionnaire is not possible incertain circumstances, for example when analyzing an existinganonymous dataset, or when filling a personality questionnaireis not convenient. APR could be used to solve this dilemma.APR is the process of mapping the data related to a subjectto a personality score that represents the personality type ofthat subject. In the context of user personality from onlinesocial network data, APR schemes are generally divided intothree classes. Text-based APR, where the source data is intext format such as social media posts or tweets. Multimedia-based APR, where the source data is an image, voice or video,such as social media profile photos. And finally, behavior-based APR, where the source data represent a set of behavioral
UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 4 patterns of the user, such as gaming behaviors or browsingbehaviors. Text-based APR generally has higher accuracy thanmultimedia-based APR and behavior-based APR.Text-based APR is inspired by the fact that some languagepsychology theories claim that the choice of words can revealsome psychological states such as emotions and personalitytraits of the subject [62]. Therefore, text-based APR analysisthe word choice frequency infer the user’s personality traitsfrom his social network posts or messages. One of the mostcommon prominent techniques for text-based APR is Linguis-tic Inquiry and Word Count (LIWC) [63], [160]. LIWC cate-gorizes the analyzed text into various psychologically relevantsets known as “buckets” like ‘function words’ (e.g., conjunc-tions, articles, pronouns), ‘social processes’ (e.g., mate, talk,friend) and ‘affective processes’ (e.g., happy, nervous, cried).Following that, LIWC measures the frequency of words ineach of these buckets and predicts the personality traits ofthe subject accordingly. Another famous linguistic databaseis the Medical Research Council (MRC) psycholinguisticsdatabase. Linguistic analysis model like LIWC and MRChave been proven to achieve acceptable accuracy to detect theuser’s personality traits from its text. For instance, Han et al. [64] introduced an APR model based on personality lexiconby analyzing the correlations between personality traits andsemantic categories of words, and extract the semantic featuresof users’ microblogs to construct a prediction model usingword classification algorithm. On the other hand, multimedia-based APR detects the user’s personality traits by analyzingit is related to photos or video and try to associate thefeatures of these data with the facets of personality traits.For instance, users who frequently post photos related toart might achieve a high score openness trait. Li et al. [65]introduced a framework that predicts the aesthetics distributionof an image and the Big-Five personality traits of people wholike the image. Finally, behavior-based APR detects the user’spersonality trait by analyzing behavioral patterns and associatethem with relevant dominant traits. Annalyn et al. [66] studiedthe relationship content labels “tags” generated by users fromGoodreads.com, and match it with personality scores collectedfrom Facebook users. Vinciarelli and Mohammadi [1] sur-veyed the literature of APR and classify the reviewed works,and Kaushal et al. [2] surveyed APR methods that leverageonline social networks as a data source. While Jacques et al. [6] surveyed vision-based APR methods, and recently, Mehta et al. [3] and Bhavya et al. [7] surveyed deep-learning-basedAPR. In Table IV, we summarize some of the key recent APRworks that were not covered in these surveys.
B. Personality matching
In personality matching phase, the recommendation systemcomputes the matching likelihood between a given user andsome items, the matching is computed based only on thepersonality information of the user and some personalityfeatures of the item, such as a product brand in productrecommendation or the personality type of actors in the case ofmovie recommendation. It is worth noting that at this stage therecommendation system does not know the rating information of the user yet, which helps the system to cope with the usercold start problem.
C. Personality filtering
The primary objective of the filtering phase in the con-ventional collaborative filtering is to determine the set ofneighbors that have similar ratings with the current user, aprocess known as neighborhood formation. In personality-aware recommendation system, the similarity between theusers is computing based on their personality trait similarity orusing a hybrid personality-rating similarity measurement, andthe resulting set of neighbors are similar in terms of personalitytraits to the studied user.III. P
ERSONALITY - AWARE SCHEMES CLASSIFICATIONS
After the emergence of personality computing, in the lastdecade, we have witnessed an unprecedented proliferationof personality-aware recommendation systems. These systemsuse different recommendation techniques, and in some cases,the recommendation process depends on the nature of therecommended content. In this section, we classify the recentpersonality-aware recommendation system based on the rec-ommendation technique. Personality-aware recommendationsystems are roughly divided into four main classes, filtering-based methods and deep-learning-based methods. Figure 3shows the classification that we will be using to classify therecent proposed personality-aware recommendation systems.Filtering methods are divided into three classes, personalityfiltering, personality matching and hybrid filtering.Fig. 3: Personality-aware recommendation systemsclassification
A. Personality filtering
Personality-aware recommendation systems that leveragethe conventional collaborative filtering technique to filter userswith similar personalities are known as personality filteringmethods. Personality filtering methods in turn could be furtherdivided into personality-neighborhood methods and matrixfactorization methods.
UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 5
TABLE IV: APR literature review
APR type Publication Year Description
Text-basedAPR Silva et al. [67] 2018 Proposed a supervised models for APR of text in Brazilian Portuguese from FacebookpostsHan et al. [64] 2020 Used word embedding techniques and prior-knowledge lexicons to automatically con-struct a Chinese semantic lexicon suitable for personality analysis.Sun et al. [68] 2020 Proposed a model of group-level personality detection by learning the influence from textgenerated networksDarliansyah etal. [69] 2019 Take advantage of Neural Network Language Model for personality detection from shorttexts by using a unified model that combines a Recurrent Neural Network named LongShort-Term Memory with a Convolutional Neural Network.Santos et al. [70] 2020 Discussed the effectiveness of using psycholinguistic knowledge in APR, and performedseries of individual experiments of APR from Facebook textMultimedia-based APR Li et al. [65] 2020 Introduced a framework that predicts the aesthetics distribution of an image and Big-Five(BF) personality traits of people who like the image.Kim et al. [71] 2018 Used computer vision techniques to detect users personality from their shared pictures. Anonline survey of 179 university students was conducted to measure user characteristics,and 25,394 photos in total were downloaded and analyzed from the respondents’Instagram accounts.Segalin et al. [72] 2016 Used Computational Aesthetics to infer the personality traits of Flickr users from theirgalleries, their method maps low-level features extracted from the pictures into numericalscores corresponding to the Big-Five Traits, both self-assessed and attributed.Ferwerda et al. [73] 2015 Conducted an online survey, by analyzing 113 participants and 22,398 extracted Instagrampictures, they conclude that there is correlation between distinct picture features andpersonality traits.Zhu et al. [74] 2020 Introduced an end-to-end weakly-supervised dual convolutional network for personalitydetection, composed of a classification network and a regression network. The clas-sification network detects personality class-specific attentive image regions. While theregression network is used for detecting personality traits.Guntuku et al. [75] 2016 Propose a personality traits detection model, and analyzed collection of images and userswho tag as favorite on Flickr.Zhang et al. [76] 2020 Proposed PersEmoN, an end-to-end trainable and deep Siamese-like network. PersEmoNis composed of two convolutional network branches, the first for emotion and the secondfor personality traits. Both networks share their bottom feature extraction module and areoptimized within a multi-task learning framework.Behavior-based APR Tadesse et al. [77] 2018 Analyzed and compared four machine learning models to investigate the relationshipbetween user behavior on Facebook and big-five personality traits.Annalyn et al. [66] 2018 Investigated the relationship between user book preferences by analyzing labels “tags”generated by users from Goodreads.com, and match it with personality scores collectedfrom Facebook usersFong et al. [78] 2015 Discussed personality inferences and intentions to befriend based solely on online avatars.Nave et al. [79] 2018 Investigated the possibility of personality prediction using musical preferences. Theirfinding using data of active listening and Facebook likes show that reactions to unfamiliarmusical excerpts predicted individual differences in personality.
1) Personality neighborhood methods:
Personality neigh-borhood filtering is the most common personality-aware rec-ommendation technique. Typically, the system uses a prox-imity function that measures the personality similarity to findthe personality neighborhood users, and use it to predict futurerating accordingly. While there are many proximity functions,the Pearson correlation coefficient is the most commonlyused proximity function. Given two users u x and u y , therating similarity between them is computed using the function SimR ( u x , u y ) as shown in eq(1), where R x and R y is thesets of the previous rating of user u x and u y respectively, and r x,i is the rating of user u x on item i , and r x is the meanrating of user u x . SimR ( u x , u y ) = (cid:80) i ∈ Rx ∩ Ry ( r x,i − r x )( r y,i − r y ) (cid:113)(cid:80) i ∈ Rx ∩ Ry ( r x,i − r x ) (cid:80) i ∈ Rx ∩ Ry ( r y,i − r y ) (1)Many personality-aware recommendation systems extendthis approach to measure the personality similarity betweenusers, as shown in eq(2), where p x and p y is the average valueof the personality traits vector for user u x and u y respectively,and p ix is the i th trait in the personality traits vector SimP ( u x , u y ) = (cid:80) i ( p ix − p x )( p iy − p y ) (cid:113)(cid:80) i ( p ix − p x ) (cid:80) i ( p iy − p y ) (2)However, some other works opted to use other proximityfunctions to measure the personality similarity between users.In Table V, we summarize some commonly used personalitysimilarity proximity functions.After computing the similarity among users and eventuallyestablishing the neighborhood of each user, the predictionscore is computed by aggregating the rating of neighbor-hood users and the similarity with these users. Formally, let score ( u, i ) denote the predication score that user u will giveto item i , the prediction score is computed using eq(3) score ( u, i ) = r u + k (cid:88) v ∈ Ω u sim ( u, v ) ( r v,i − r v ) (3)where r u and r v are the average rating of user u and user v respectively, and r v,i is the rating given by user v to item i ,and Ω u are the neighbors of user u that have previously rateditem i . Different works used a different design of the proximity UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 6 function that measures the total similarity sim ( u, v ) . In thisregard, there are three main designs. Some works ([80]) simplyuse the personality similarity function simp ( u, v ) instead of sim ( u, v ) . While other works ([57]) opted to use a combi-nation of the personality similarity function simp ( u, v ) andthe rating similarity function simr ( u, v ) . Finally, some otherworks ([81], [33]) use other social factor similarity functionssuch as user interests similarity along with the rating similarityand personality similarity.
2) Matrix factorization methods:
In personality-enhancedmatrix factorization methods, the conventional matrix fac-torization algorithm is extended to incorporate the user’spersonality traits along with its ratings. In the conventionalmatrix factorization method, the user-item interaction matrixis decomposed to the product of two low-dimensionality rect-angular matrices that represent the represent users and itemsin a lower dimensional latent space, this is done by applyingdimensionality reduction algorithm such as singular valuedecomposition. Formally, let p u ∈ R k and q i ∈ R k denotethe latent feature vector of user u and item i respectively.In the conventional matrix factorization method, the user u’spreference to item i is estimated by computing the dot productof user u and item i latent feature vectors, as shown in eq(4). ˆ r ui = p u .q i (4)Personality-enhanced matrix factorization extends this byincorporating the user’s personality traits, and other occasion-ally other social attributes. They introduce an additional latentfeature vector y a ∈ R k for each social attribute a ∈ A .eq(4) is extended to incorporate these attributes as shown ineq(5). It is worth noting that some works consider the Big-Five personality score vector as the only attribute, as in [59].While some other works add other social attributes in additionto the personality, for instance, in [30] the user’s gender, agegroup and the scores for the Big Five personality traits areused as attributes. ˆ r ui = q i . (cid:32) p u + (cid:88) a ∈ A y a (cid:33) (5) B. Personality matching methods
This approach is similar to the conventional content filteringapproach. In the personality matching method, a personalityscore is assigned to each item. This is done either usingcontent analysis, attribute analysis or a hybrid approach. Inthe content analysis, the system assigns a personality scoreto an item by applying an automatic personality recognitionon the content of that item, such as the textual description,labels and category. For example, in [26] a product personalityassessment method known as product personality scale [90]was used to assess the personality of the items. While inattribute analysis, the system assigns a personality score toan item by analyzing the attribute of that item, such as thepersonality traits of users that interacted with that item. Forinstance, in [86] the personality P G i of a video game G i isassigned by computing the average personality traits of users who played G i as P G i = (cid:80) Uj ∈ OGi P Uj | O Gi | . Personality matchingis usually applied if we can observe a common matchingcriteria between the recommended content and the target user,which eliminate the need for extensive computing in order tofind the neighborhood set from one hand, and mitigate thecold start on the other hand. Xiao et al. [91] used personalitymatching approach for followee recommendation, the totalpersonality matching (TPM) score between a given user u andthe potential blogger pf is computed as shown in eq(6): T P M ( u, pf ) = µ ( M S ( u, pf, dim )) (6)Where M S ( u, pf, dim ) denotes the personality matchingscore of the user u and the potential recommendation followee pf in a the respective dimension, and µ is the average valueof each dimension. C. Hybrid personality filtering methods
Hybrid personality filtering methods combine the techniqueof personality filtering on the users’ space, and personalitymatching on the items space. Hybrid personality filteringhas been proven as an effective method that leverages theadvantages of personality filtering and personality matchingmethods. [57], [92] used a hybrid personality filtering ap-proach for friend recommendation, where personality filteringis used to find users with similar ratings and personalitymatching is used to filter the item space (since it is a friendrecommendation system, the items represent potential friends).Similarly, [86] also used hybrid personality filtering for agame recommendation, where personality filtering is used todetermine user with similar game ratings, and personalitymatching is used to attribute personality to games.
D. Deep learning methods
In recent years, deep learning has revolutionized the domainof recommendation systems by leveraging deep learning mod-els, such as Convolutional Neural Network (CNN), RecurrentNeural Network (RNN) and Autoencoder, to name a few.Personality-aware recommendation systems are not an excep-tion to this revolution. Deep learning is either used to detect theuser personality of the users or in the recommendation processitself. The choice of deep learning model used for personalitydetection and personality-aware recommendation depends onthe type of source data of users. Deep learning model thatare inspired by natural language processing, such as the n-gram model, are suitable for personality detection and contentrecommendation from textual source data. For instance, Ma-jumder et al. [107] proposed deep CNN for document-level au-tomatic personality recognition, the CNN extracts monogram,bigram and trigram features from the document text and eachword was represented in the input as a fixed-length featurevector using Word2Vec [108] model, finally linguistic features(e.g. LIWC, MRC) are concentrated and fed to fully connectedlayer for personality traits prediction. Similarly, deep learningmodels that are designed for image and video processingare suitable for personality detection and personality-awarerecommendation using non-textual personality data. Wei et al.
UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 7
TABLE V: Personality-based similarity measurement
Publication Proximity function Formula Note [82], [57],[83], [84],[47], [32],[34], [85],[31] Pearson correlationcoefficient
SimP ( u, v ) = (cid:80) i ( p iu − p u )( p iv − p v ) (cid:113)(cid:80) i ( p iu − p u ) (cid:80) i ( p iv − p v ) Used by most of the state-of-the-artpersonality-aware recommendation systems[41], [36] Normalized Euclideandistance
SimP ( u, v ) = (cid:18)(cid:113)(cid:80) k w k ( P ku − P kv ) (cid:19) w k is the weight of the k th personality trait.[26] Euclidean distance Eucl = (cid:113) ( E BFI − S − E PPS ) +( A BFI − S − A PPS ) +( C BFI − S − C PPS ) E BFI − S is the extraversion trait value forthe user, and E PPS is the extraversion traitvalue assigned to a product. Similarly A isthe agreeableness trait, and C is conscien-tiousness[86], [80],[25], [87] Cosine similaritymeasure
SimP ( u, v ) = (cid:80) k P u,k .P v,k √ (cid:80) k P u,k √ (cid:80) k P v,k Cosine similarity is a dot product of unitvectors, unlike Pearson correlation, which iscosine similarity between centered vectors.[88] Spearman’scorrelation coefficient
SimP ( u, v ) = (cid:80) k ( s u,k − s u )( s v,k − s v ) (cid:113)(cid:80) k ( s u,k − s u ) (cid:80) k ( s v,k − s v ) Where s u,k is the position of P u,k in thedecreasing order ranking of Big Five scores.[89] Hellinger-Bhattacharya Distance SimP ( u, v ) = √ ∗ (cid:113)(cid:80) i =1 (cid:0) √ u i − √ v i (cid:1) TABLE VI: Personality-aware recommendation techniquesclassification
Recommendation method Publication
Personality neighborhood [80], [93], [94], [82],[57], [83], [84], [47],[32], [41], [26], [86],[80], [88], [89] [95],[96], [31]Matrix factorization [30], [59], [97], [98],[99], [25], [34], [100],[101]Personality matching [26], [102], [103],[104] [91]Hybrid personality filtering [57], [33] [86], [81],[105]Deep-learning [106], [107], [108],[109], [110], [39],[111], [112], [65],[110], [113] [109] proposed Deep Bimodal Regression (DBR) frameworkfor apparent personality analysis from videos and images.DBR modified the common convolutional neural networks forincorporating essential visual cues. Besides the source dataformat, the recommended content nature could also influencethe choice of deep learning models to be used as personality-aware recommendations. For example, Chi-Seo et al. [110]introduced a system that employs deep learning to classify andrecommend tourism types that are compatible with the user’spersonality. The model is composed out of three layers, eachlayer incorporates a service provisioning layer that real usersface, the recommendation service layer, responsible to producerecommended services based on user information inputted, andthe adaptive definition layer, that learns the types of tourismthat fits for the user’s personality types.In Table VI we classify some of the recent personality-awarerecommendation systems based on their recommendation tech-nique. IV. L
ITERATURE REVIEW OF PERSONALITY - AWARERECOMMENDATION SYSTEMS
In the last few years, we have witnessed a rapid proliferationof personality-aware recommendation systems. In this section,we review the literature on personality-aware recommendationsystems in different application domains.
A. Friend recommendations
In the literature of social networks, many personality-awarefriend recommendation systems have been proposed, Ning etal. [57] proposed a personality-aware friend recommendationsystem named PersoNet that leverages Big-Five personalitytraits to enhance the hybrid filtering friend selection process.PersoNet outperformed the conventional rating-based hybridfiltering, and achieve acceptable precision and recall valuesin cold start phase as well. Similarly, Chakrabarty et al. [89] designed a personality-aware friend recommendation sys-tem name FAFinder (Friend Affinity Finder). FAFinder usesHellinger-Bhattacharyya Distance (H-B Distance) to measurethe user’s Big-Five similarity and recommend friends accord-ingly. While Bian et al. [102] designed and implementedMatchmaker, a personality-aware friend recommendation sys-tem that recommends friends to users on Facebook by match-ing and comparing user’s online profile with the profiles of TVcharacters. For example, if Facebook user X is similar to TVcharacter 1, and Facebook user Y is similar to TV character2, and character 1 and character 2 are friends in the sameTV show, then the Matchmaker system recommends user Xto become friends with user Y. Whereas, Neehal et al. [39]introduced a personality-aware friend recommendation frame-work, which uses a 3-Layered artificial neural network (ANN)for friend preference classification and a distance-based sortedsubset selection function for friend recommendation.Tommasel et al. [114], [115] studied the effects of userpersonality on the accuracy of followees prediction on mi-croblogging social media. The authors analysed how the user’spersonality character influence the followees selection process
UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 8 by incorporating personality traits with state of the art followeepredicting factors. To prove the effectiveness of proposedfollowee prediction algorithm, the author collected a Twitterdataset by crawling the account of 1852 users, and only usersthat English is their tweeting language were selected. Theytested the content-based followee prediction algorithm withand without including the user’s personality traits. Their resultsshowed that incorporating personality traits can enhance thefollowee recommendations. While in Tommasel et al. [116],they analyzed 3 different similarity factors. Firstly, they cal-culated the total similarity by taking into account the BigFive personality factors as a whole. Secondly, they calculatedthe dimension to dimension similarity measure by taking intoaccount every individual personality traits separated from eachother. Finally, they calculated a cross dimension similaritymeasurement by taking into account every personality facedin relation to the others. Their results showed that personalitytraits must be regarded as a distinctive factor in the process offollowee prediction. However, personality dimensions shouldnot be analyzed as a whole because the overall personalitysimilarity measurement might not precisely assess the actualmatching between users. The data analysis proves the exis-tence of relations among the individual personality facades.Therefore, the importance of assessing each personality traitwith respect to other users. Similarly, Xiao et al. [91] intro-duced a personality-aware followee recommendation systembased on sentiment analysis and text semantics, they proposedmodel combines the user attributes with the Big-Five traits torecommend new followees. And Mukta et al. [58] proposeda technique to detect homophily by analyzing the Big-Fivepersonality traits of users in an egocentric network such asFacebook.
B. Movies recommendations
Asabere et al. [82] proposed ROPPSA, a personality-aware TV program recommendation system that leveragesnormalization and folksonomy procedures to generate grouprecommendations for viewers with similar personality traits.Balakrishnan et al. [42] proposed a hybrid recommendersystem for movies named HyPeRM,, the proposed systemincludes the users’ personality character in addition to theirdemographic information (e.g. sex and age) to enhance theprecision of the recommendations. Big Five personality traitwas employed to measure the users’ personalities. HyPeRMwas tested based on the Root Mean Square Error of Approx-imation (RMSEA) and the Standardized Root Mean SquareResidual (SRMR). Both theses metrics showed that HyPeRMoutperformed the baseline variant (i.e. the recommendationwithout including the user’s personality) in terms of theprecision of the recommendation. Their work shows thatmovies recommendations can be improved by incorporatingthe viewers’ personality traits. Similarly, Shanchez et al. [117]proposed HappyMovie, a Facebook application for movie rec-ommendations, HappyMovie uses three features for a movierecommendation, user personality, social trust with other usersand the past movie ratings. While Bolock et al. [118] pro-posed a movie recommendation system based on the user’s character. The system is adaptive in the way it uses a differentrecommendation algorithm for different users based on theused character criteria. The authors implemented a movierecommendation application to find the relationship betweenthe user’s character and the recommendation algorithm, theyhave used three main character dimensions, user personality,background and gender.On the relationship between personality and movie prefer-ences. Golbeck et al. [119] proved the positive correlationbetween personality and users’ movie preferences. Usingsurveys and analysis of system data for 73 Netflix users,they proved correlations between personality and preferencesfor specific movie genres. Wu and Chen [80] studied userpersonality inferences using implicit behaviors with movies,and the possibility to recommend movies base on the user’spersonality traits without users’ explicit ratings. Specifically,they determined a set of behavioral features using experimentalconfirmation and proposed an inference method using Gaus-sian Process to fuse these features and subsequently detectthe user’s Big-Five personality traits. After that, they used theobtained personality information to enhance the collaborativefiltering movie recommendation process. Scott et al. [120]investigated the relationship between personality and culturaltraits with a perception of multimedia quality. Hu et al. [32] addressed the cold-start problem by incorporating humanpersonality into the collaborative filtering framework, theyhave tested the proposed system with movies and music publicdatasets. And Berkovsky et al. [28], studied the effects of dif-ferent recommendation and content filtering strategies on usertrust. They evaluated the score of nine main factors of trustgrouped and divided them into three dimensions and tested thedifferent observations regarding the users’ personality traits.
C. Music recommendations
The authors of the works presented in [121], [122], [123]discussed the impact of personality traits on the accuracy ofmusic recommendation systems. Cheng et al. [103] Introduceda hybrid method for personality-aware music recommenda-tions. They used personality matching of the user’s personalitytraits with an extracted feature from songs audio, and classifythese features using support vector machine (SVM) algorithm.While Schedl et al. [43] studied the relationship betweenpersonality traits and classical music preferences, they groupedthe users into four clusters based on the personality traits andtried to infer the preference of each cluster regarding classicalmusic. Ferwerda et al. [124] discussed the possibility toenhance music recommendation systems by incorporating theuser’s psychological factors such as emotional and personalitystates. The authors discussed how people listen to music tocontrol their emotional states, and how this adjustment isrelated to their personality traits. They focused on the methodsto acquire data from social media networks to estimate thecurrent emotional state of the listeners. Finally, they discussedthe connection of the accurate emotionally with the musiccategories to support the emotional adjustment of listeners.The same research group proposed a personality-aware musicrecommender system [123], where they employed the users’
UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 9 personality traits as a general model. The authors specified therelationships between listeners’ personality and their behavior,preferences, and needs, in addition to that the authors studiedthe different ways to infer users’ personality traits from user-generated data of social media websites (e.g., Facebook,Twitter, and Instagram). Hu et al. [29] proposed a generalmodel that can deduce users’ music preferences based ontheir personality characteristics. Their subject studies provethat most of the active users think that the recommended songsare more precise for their friends, however, these users enjoymore using personality questionnaires based recommendersfor finding songs for themselves. The authors investigate ifdomain based knowledge has an impact on users’ under-standing of the system. We found that novice users, whoare less knowledgeable about music, generally appreciatedmore personality based recommenders. Zhou et al. [93] useddecision trees to developed a heuristic personality-aware musicrecommendation system for niche market.
D. Image recommendation
Associating images features and personality traits istwofold: known the feature of the image can help to in-fer the personality of users who interact with the image,and know the personality traits of the users could help torecommend relevant images. Guntuku et al. [55] studiedmethods for modeling the personality character of users basedon a collection of images that they tagged as ‘favorite’ onFlickr. Their study presents several methods for enhancingpersonality detection performance by proposing better featuresand modeling approaches. They evaluated their approach bymeasuring its efficiency when used in an image recommen-dation system. The presented results showed the need forusing high-level user understandable features and illustratethe effectiveness of a +A2P (Answers-to-Personality) and F2A(Features-to-Answers) approaches compared to the traditionalF2P (Features-to-Personality) method that is usually used byexisting works. While in [125] they studied the effects ofpersonality (Big-Five Model) and cultural traits (HofstedeModel) on the potency of multimedia-stimulated positive andnegative emotions. Wu et al. [126] proposed a hierarchicalattention model for social feature based image recommenda-tion, in which the recommendation system considers the socialcharacteristics of the users, such as user social interests andpersonality traits. Li et al. [112] developed a deep-learning-based image aesthetic model that employs the user’s person-ality traits for image aesthetic rating. The personality featuresare used to represent the aesthetics features, hence, producingthe optimal generic image aesthetics scores. Furthermore, in[65] they extended their method to offer a personality-awaremulti-task framework for generic as well as personalizedimage aesthetics assessment. Gelli et al. [127] investigatedthe effects of personality on user behaviors with images ina social media, and which visual stimuli contained in photocontent can affect user behaviors. They analyzed a twitterdataset of 1.6 million user and image retweet behaviors. Kim et al. [128] studied the relationships between Instagram userpersonality traits and color features of their photos, and found that agreeableness is the most relevant trait that is associatedwith the photo and color features.
E. Academic content recommendations
Many works have used personality traits for academic-oriented recommendation systems, such as courses recom-mendations, conference attendee recommendations and re-search paper recommendations. Xie el al [83] proposed arecommendation system of academic conference participantscalled SPARP (Socially-Personality-Aware-Recommendation-of-Participants). For more effective collaborations in the visionof a smart conference, the proposed recommendation approachuses a hybrid model of interpersonal relationships amongacademic conference participants and their personality traits.At first, the proposed system determines the social ties amongthe participants based on past and present social ties fromthe dataset with four trial-weight parameters. These weightparameters are used later in their experiment to representvarious influence proportions of the past and present socialties among participants. Following that, the system calculatesthe personality-similarity between the conference participantsbased on explicit tagged-data of the personality ratings. Sim-ilarly, Asabere et al. [84] proposed a recommendation algo-rithm for conference attendees called PerSAR (Personality-Socially-Aware-Recommender). The proposed system is basedon a hybrid approach of social relations and personalitycharacters of the conference participants. To evaluate theirproposed system, the authors used the dataset of The Inter-national Conference on Web-Based Learning (ICWL) 2012,which includes the social ties of 78 conference participantswith a total time-frame of 12 hours (720 minutes). Far fromthat, Fahim Uddin et al. [129] Proposed a personality-awareframework to improve academic choice for newly enrolledstudents. Their proposed framework makes use of the researchfield of Predicting Educational Relevance For an EfficientClassification of Talent, which uses stochastic probabilitydistribution modeling to help the student to choose the relevantacademic field. Hariadi et el [94] proposed a personality-aware book recommendation system that combines the user’sattributes as well as his personality traits. The proposedsystem leverages MSV-MSL (Most Similar Visited Material tothe Most Similar Learner) method to compute the similaritybetween users and form the personality neighborhood.
F. Product recommendations
Dhelim et al. [81] introduced Meta-Interest, a personality-aware product recommendation system that considers theuser interests and personality traits and recommends rele-vant products by exploring the possible user-item metapaths.Tkalcic et al. [36] proposed a new approach for measuringthe user similarity for collaborative filtering recommendersystems that is based on the Big Five personality modelin the context of product recommendation. Buettner [26]introduced a personality-aware framework for product recom-mender named PBPR. The proposed framework analyzes theuser’s social media profile to infer its personality traits andrecommend products accordingly. The author evaluated his
UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 10 proposed framework as IT artefact using a dataset from XING.Huang et al. [130], used a data driven method to predict onlineshoppers’ online buying preferences. Firstly, the authors usedtext mining method based on the shoppers’ language usage be-haviors to create seven different dimension lifestyle-lexicons.Following that, they included these lifestyle-lexicons in theproduct recommendation system that can predict the shoppers’buying preferences. Roffo [54] discussed utilizing personalityto compute the association between the shopper’s purchasingtendency and the advert’s recommendations. Moreover, theauthor introduced the ADS dataset, an advertising benchmarkenriched with Big Five personality traits of users along with1200 personal photos. Adamopoulos and Todri [131] useda dataset from Amazon.com to evaluate a personality basedrecommendation, they have inferred the users’ personalitytraits along with their needs and other contextual informationfrom their social media profiles. Their findings is that addingpersonality to the recommendation process can increase theefficiency of the system.
G. Game recommendations
Yang et al. [86] introduced a personality-aware game rec-ommendation system, they apply text mining on the players’social network posts to extract their personality types andanalyzed the games’ content to associate these games withcertain personality types. They proved the effectiveness oftheir proposed system through an experiment on 63 playersand more than 2000 games. Lima et al. [132] designed anew method for interactive storytelling in games, in whichthe quests and the ongoing story follow the view of individualpersonality traits and behaviors in a non-deterministic way.Chan et al. [133] proposed a method for matching playersusing personality types to augment the enjoyment and socialinteraction in exergames. They argue that maintaining highlevels of enjoyment and active social interactions is crucialbecause both can offer retention and continuation of game-play and exercise involvement. Hill et al. [134] investigatedthe association between HEXACO personality model withpreferences for certain aspects of gaming experiences. Themain finding confirmed that extraversion trait is moderatelyassociated with the socializer gaming preference and a slightassociation with the daredevil gaming preference. While Ab-basi et al. [135] discussed the personality differences betweengamers and non-gamers. Supported by evidence obtained byanalyzing the personality types of 855 students (gamers andnon-gamers), they conclude that gamers have a personalitytypes that is significantly different on compared to non-gamers.
H. Points of interest recommendations
Wang et al. [136] proposed a trust-based POI recommenda-tion system, they leverage the personality similarity betweenusers to compute the trust level. In addition to trust andpersonality information, they also make use of the graphicand temporal influence in the recommendation model. Chi-Seo et al. [110] introduced a system that employs deep learningto classify and recommend tourism types that are compatiblewith the user’s personality. The model is composed out of three TABLE VII: Personality-aware personality recommendationsystem content-based classification
Recommendationdomain Recommendation system
Friend recommendation [57] [58], [89], [102], [39], [114],[115], [116], [91], [139]Image recommendation [55], [111], [125], [126], [112],[65], [127], [128], [36]Movie recommendation [28], [42], [80], [117], [118],[119], [140], [100], [120], [32],[40], [85], [141], [45], [142],[143], [144], [145], [146]Music recommendation [106], [147], [43], [93], [103],[121], [122], [123], [124], [29],[46], [87], [148], [149], [150],[151], [152], [153], [35], [154]Product recommendation [26], [97], [54], [36], [130],[131], [56], [155], [31], [101]POI recommendation [110], [98], [99], [113], [136],[138], [104], [137], [61], [156],[157], [158], [159]Academic recommenda-tion [25], [83], [84], [129], [94],[129], [161], [162]Game recommendation [86], [132], [133], [134], [135],[163] layers, each layer incorporates a service provisioning layer thatreal users face, the recommendation service layer, responsibleto produce recommended services based on user informationinputted, and the adaptive definition layer, that learns the typesof tourism that fit for the user’s personality types. Zhang et al. [113] introduced a new POI recommendation systemthat uses POI classification model named POIC-ELM. POIC-ELM extracts 9 features that are related to 3 factors, theuser’s personality information the POI information and theuser’s social relationships information. The learned featureare then fed to an extreme learning machine (ELM) for POIclassification. Braunhofer et al. [98] introduced STS (SouthTyrol Suggests), a personality-aware POI recommender systemthat uses an active learning module and personality-awarematrix factorization recommendation to infer the relevant POI.In the same vein, in [137] they designed a personalizedactive learning method that takes advantage of the user’spersonality information to get more accurate in-context POIratings. Tanasescu et al. [104] introduced the concept of’personality of a venue’. They extracted keywords and otherannotations from the reviews of the venues and mapped theseinformation to Big-Five personality traits. The experimentaltesting confirmed the correlation between visitors’ personalitytraits and the personality of the visited venue. Sertkan et al. [138] proposed an automatic method for computing the Seven-Factor equivalent of tourism sites. Regression analysis, clusteranalysis, and exploratory data analysis are performed to findthe correlation between Seven-Factors and the type of touristdestination site. Feng et al. [99] fused three factors, mainlyinterpersonal interest similarity, personal interest similarityand interpersonal influence to implement probabilistic matrixfactorization for personality-aware recommendations.In Table VII, we summarize the reviewed works relatedto personality-aware recommendation systems based on therecommended content.
UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 11
V. D
ATASETS AND B ENCHMARKS
Due to the availability of open public datasets that con-sidered the users’ personality information, many personality-aware recommendation systems were able to train their pro-posed models and compare them using the state-of-the-artbenchmarks. In this section, we present two of the widelyused personality datasets in the context of personality-awarerecommendation systems.
A. myPersonality dataset
In 2007, David Stillwell a PhD student at the University ofNottingham designed a Facebook application called myPer-sonality that leverages IPIP version of the NEO personalityinventory personality questionnaire and displays the personal-ity score instantly [164]. myPersonality was initially intendedfor limited use, David shared it with his close friends. Later on,surprisingly the number of users who joined the study increasedramatically, and many users were willing to donate their datato be used for academic purposes. By 2012, more than 6million users finished the IPIP personality questionnaire, andthe respondents came from different age groups, backgroundsand cultures. myPersonality dataset was anonymized and sam-ples of it were shared with many researchers. In 2018, thecreators of myPersonality decided to stop the project, as it hasbecome extremely challenging to maintain the dataset with theincreasing number of usage requests from researchers over thelast few years.
B. MovieLens dataset
MovieLens is a widely used open dataset in recommen-dation system researches. It contains movie rating data ex-tracted from the famous movie recommendation and ratingwebsite MovieLens.com [165]. The ratings were collected overdifferent periods of time, there are many available versionsof the dataset depending on the size of the dataset. Thelargest available version of the dataset is named MovieLens25M. It contains 25 million movie ratings and one milliontag applications applied to 62,000 movies by 162,000 users.Personality2018 [24] is a version of MovieLens dataset thatincludes the personality information of the users that rated themovies, it contains the TIPI score of 1834 users along withthe movie rating that were given by these users.
C. Newsfullness dataset
Newsfullness is a news sharing platform that usespersonality-aware recommendation for of news articles [33].Newsfullness contains more the TIPI score of 2228 users alongwith their articles that these users viewed or liked. The col-lected articles were from all the main news websites, such asBBC, CNN, RA and Aljazeera, from different news categories(business, politics, health, sports, travel, entertainment, art,education, science and technology). Table VIII summarizesthe works that have used these datasets. TABLE VIII: Personality datasets
Dataset Works Content myPersonality [34], [95], [166], [87], [100], [148],[161], [106] N/AMovieLens [167], [40], [85], [96], [140], [32],[149], [31], [93] MoviesNewsfullness [33], [81], [105] News articlesTwitter [145], [161], [114], [116], [115] FriendshipsIMDB [139], [28], [145] MoviesLast.fm [87], [148], [151], [153] MusicOther datasets allrecipes.com [168], Amazon[169],Steam [163], [86] ADS [54], [56],Douban [80], [143], [170], IAPS[36],Facebook [58], , PsychoFlickr [55],Sina Weibo[91], Deezer [106] N/A
VI. C
HALLENGES AND OPEN ISSUES
Although that personality-aware recommendation systemoffers many advantages and solutions to tackle recommen-dation challenges that conventional recommendation systemscannot solve, such as cold start and recommendation diversity.However, using the user’s personality in the recommendationbring up new challenges and ethical issues, in this section wediscuss some of these challenges.
A. Personality information privacy
The privacy of the user’s personality poses a new challengein addition to the existing challenge of preserving the privacyof user’s information. As the user’s personality information iseven more sensitive than other information in the user’s profile.In March 2018, Facebook-Cambridge Analytica scandal hasdrawn the attention of the world. A Facebook applicationcreated by the a data analytic company named CambridgeAnalytica unrightfully collected the personality informationof more than 87 million users, aiming to manipulate theirvoting choice in the 2016 US presidential election [171]. Thechallenge of personality-aware recommendation system is topreserve the personality information of the users, as malicioususers can analyze the recommendation results to predict thepersonality type of other users. The recommendation systemis responsible to maintain the tradeoff between data sharingand information privacy between users.
B. Measurement accuracy
The accuracy of the personality measurement is vitalfor personality-aware recommendation system, the inaccuratemeasurement of the user personality traits will inevitably leadto inaccurate recommendations. What makes things worseis that the system considers personality traits as contentinformation that do not need update frequently, and willoffer inaccurate information all the time. The personalityquestionnaire contains questions that are relative to the subjectitself, and there is no standard measurement of the questionedfeatures, which could increase the reference-group effect. Forinstance, an introverted subject may identify himself as anextrovert, even if he filled the questionnaire correctly, thatis because all his close friends are also introverts, thereforehis judgment was relative to his environment. APR methodsmay also inaccurately detect the user’s personality for various
UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 12 reasons, for example, image-based APR might predict thepersonality of a user by analyzing his shared photos on socialmedia without considering the context of these photos. Forexample, a user who shares nature photos frequently as apart of his job as a photographer or a war photo shared bya journalist may not reflect their personalities.VII. C
ONCLUSION
To the best of our knowledge, this survey is the first thatfocuses on personality-aware recommendation system. Wehave reviewed the literature of the recent works in this domain,and show the main differences between different works, interms of personality model, as well as in terms of the usedrecommendation technique. The vast majority of personality-aware recommendation systems leverage Big-Five personalitymodel to represent the user’s personality. Personality-awarerecommendation systems have the upper hand when comparedwith the conventional recommendation techniques, especiallywhen dealing with cold start and data sparsity problems.However, with the understanding of the user’s personalityadvantage comes the challenge of preserving the privacy ofthe user personality information, and also the challenge ofmaintaining a high personality detection accuracy.A
CKNOWLEDGMENTS
We would like to thank all the active users of Newsfullnessthat agreed to be a part of the Meta-Interest experiment.This work was supported by the National Natural ScienceFoundation of China under Grant 61872038, and in part bythe Fundamental Research Funds for the Central Universitiesunder Grant FRF-BD-18-016A.R
EFERENCES[1] A. Vinciarelli and G. Mohammadi, “A Survey of PersonalityComputing,”
IEEE Transactions on Affective Computing , vol. 5, no. 3,pp. 273–291, jul 2014. [Online]. Available: http://ieeexplore.ieee.org/document/6834774/[2] V. Kaushal and M. Patwardhan, “Emerging Trends in PersonalityIdentification Using Online Social Networks—A Literature Survey,”
ACM Transactions on Knowledge Discovery from Data , vol. 12, no. 2,pp. 1–30, mar 2018. [Online]. Available: https://dl.acm.org/doi/10.1145/3070645[3] Y. Mehta, N. Majumder, A. Gelbukh, and E. Cambria, “Recent trendsin deep learning based personality detection,”
Artificial IntelligenceReview , vol. 53, no. 4, pp. 2313–2339, apr 2020. [Online]. Available:http://link.springer.com/10.1007/s10462-019-09770-z[4] A. G.C. Wright, “Current Directions in Personality Science and thePotential for Advances through Computing,”
IEEE Transactions onAffective Computing , vol. 5, no. 3, pp. 292–296, jul 2014. [Online].Available: http://ieeexplore.ieee.org/document/6933966/[5] A. Vinciarelli and G. Mohammadi, “More Personality in PersonalityComputing,”
IEEE Transactions on Affective Computing , vol. 5, no. 3,pp. 297–300, jul 2014. [Online]. Available: http://ieeexplore.ieee.org/document/6866147/[6] J. C. Silveira Jacques Junior, Y. Gucluturk, M. Perez, U. Guclu,C. Andujar, X. Baro, H. J. Escalante, I. Guyon, M. A. J. Van Gerven,R. Van Lier, and S. Escalera, “First Impressions: A Survey on Vision-based Apparent Personality Trait Analysis,”
IEEE Transactions onAffective Computing , 2019.[7] S. Bhavya, A. S. Pillai, and G. Guazzaroni, “Personality identificationfrom social media using deep learning: A review,” in
Soft Computingfor Problem Solving . Springer, 2020, pp. 523–534. [8] S. V. Kedar and D. S. Bormane, “Automatic personality assessment: Asystematic review,” in . IEEE, dec 2015, pp. 326–331. [Online]. Available:http://ieeexplore.ieee.org/document/7489402/[9] A. N. Finnerty, B. Lepri, and F. Pianesi,
Acquisition of Personality .Cham: Springer International Publishing, 2016, pp. 81–99. [Online].Available: https://doi.org/10.1007/978-3-319-31413-6 { }
Personality and Individual Differences , vol. 124, pp.150–159, apr 2018. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/S0191886917307328[11] P. S. Dandannavar, S. R. Mangalwede, and P. M. Kulkarni, “SocialMedia Text - A Source for Personality Prediction,” in . IEEE, dec 2018, pp. 62–65.[Online]. Available: https://ieeexplore.ieee.org/document/8769304/[12] T. Santamaria and D. Nathan-Roberts, “Personality Measurementand Design in Human-Robot Interaction: A Systematic and CriticalReview,”
Proceedings of the Human Factors and ErgonomicsSociety Annual Meeting , vol. 61, no. 1, pp. 853–857, sep2017. [Online]. Available: http://journals.sagepub.com/doi/10.1177/1541931213601686[13] L. Robert, “Personality in the human robot interaction literature:A review and brief critique,” in
Proceedings of the 24th AmericasConference on Information Systems, Aug , 2018, pp. 16–18.[14] L. Robert, R. Alahmad, C. Esterwood, S. Kim, S. You, and Q. Zhang,“A Review of Personality in Human–Robot Interactions,” { SSRN } Electronic Journal , 2020. [Online]. Available: https://doi.org/10.2139/ssrn.3528496[15] Y. Mou, C. Shi, T. Shen, and K. Xu, “A Systematic Reviewof the Personality of Robot: Mapping Its Conceptualization,Operationalization, Contextualization and Effects,”
InternationalJournal of Human–Computer Interaction
Personalityand Individual Differences , vol. 38, no. 3, pp. 591–604, feb2005. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/S0191886904001394[17] P. T. Costa Jr and R. R. McCrae,
The Revised NEO PersonalityInventory (NEO-PI-R).
Sage Publications, Inc, 2008.[18] B. Rammstedt and O. P. John, “Measuring personality in oneminute or less: A 10-item short version of the Big Five Inventoryin English and German,”
Journal of Research in Personality ,vol. 41, no. 1, pp. 203–212, feb 2007. [Online]. Available:https://linkinghub.elsevier.com/retrieve/pii/S0092656606000195[19] S. D. Gosling, P. J. Rentfrow, and W. B. Swann, “A very briefmeasure of the Big-Five personality domains,”
Journal of Research inPersonality , vol. 37, no. 6, pp. 504–528, dec 2003. [Online]. Available:https://linkinghub.elsevier.com/retrieve/pii/S0092656603000461[20] E. Topolewska, E. Skimina, W. Strus, J. Cieciuch, and T. Rowi´nski,“The short IPIP-BFM-20 questionnaire for measuring the Big Five,”
Roczniki Psychologiczne , vol. 17, no. 2, pp. 385–402, 2014.[21] C. A. Pedregon, R. L. Farley, A. Davis, J. M. Wood, andR. D. Clark, “Social desirability, personality questionnaires, andthe “better than average” effect,”
Personality and IndividualDifferences
Psychological science , vol. 28, no. 3, pp. 276–284, 2017.[23] R. P. Karumur, T. T. Nguyen, and J. A. Konstan, “Personality, UserPreferences and Behavior in Recommender systems,”
InformationSystems Frontiers , vol. 20, no. 6, pp. 1241–1265, dec 2018. [Online].Available: http://link.springer.com/10.1007/s10796-017-9800-0[24] T. T. Nguyen, F. Maxwell Harper, L. Terveen, and J. A.Konstan, “User Personality and User Satisfaction with RecommenderSystems,”
Information Systems Frontiers , vol. 20, no. 6, pp. 1173–1189, dec 2018. [Online]. Available: http://link.springer.com/10.1007/s10796-017-9782-y[25] Y. Zheng and A. Subramaniyan, “Personality-aware collaborativelearning: Models and explanations,” in
International Conference onAdvanced Information Networking and Applications . Springer, 2019,
UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 13 pp. 631–642. [Online]. Available: http://link.springer.com/10.1007/978-3-030-15032-7 { }
ElectronicMarkets , vol. 27, no. 3, pp. 247–265, aug 2017. [Online]. Available:http://link.springer.com/10.1007/s12525-016-0228-z[27] W. Wu, L. Chen, Q. Yang, and Y. Li, “Inferring Students’Personality from Their Communication Behavior in Web-basedLearning Systems,”
International Journal of Artificial Intelligence inEducation , vol. 29, no. 2, pp. 189–216, may 2019. [Online]. Available:http://link.springer.com/10.1007/s40593-018-00173-9[28] S. Berkovsky, R. Taib, and D. Conway, “How to recommend? Usertrust factors in movie recommender systems,” in
Proceedings of the22nd International Conference on Intelligent User Interfaces , 2017,pp. 287–300.[29] R. Hu and P. Pu, “A study on user perception of personality-basedrecommender systems,” in
International conference on user modeling,adaptation, and personalization . Springer, 2010, pp. 291–302.[30] M. Elahi, M. Braunhofer, F. Ricci, and M. Tkalcic, “Personality-basedactive learning for collaborative filtering recommender systems,” in
Congress of the Italian Association for Artificial Intelligence . Springer,2013, pp. 360–371.[31] R. Hu and P. Pu, “Using personality information in collaborativefiltering for new users,”
Recommender Systems and the Social Web ,vol. 17, 2010.[32] ——, “Enhancing collaborative filtering systems with personalityinformation,” in
Proceedings of the fifth ACM conference onRecommender systems - RecSys ’11 . New York, New York,USA: ACM Press, 2011, p. 197. [Online]. Available: http://dl.acm.org/citation.cfm?doid=2043932.2043969[33] S. Dhelim, N. Aung, and H. Ning, “Mining user interestbased on personality-aware hybrid filtering in social networks,”
Knowledge-Based Systems , p. 106227, jul 2020. [Online]. Available:https://linkinghub.elsevier.com/retrieve/pii/S0950705120304354[34] J. Sun, D. Ren, and D. Xu, “Leveraging User Personality and TagInformation for One Class Collaborative Filtering,” in
Advances inMultimedia Information Processing – PCM 2018 , R. Hong, W.-H.Cheng, T. Yamasaki, M. Wang, and C.-W. Ngo, Eds. Cham: SpringerInternational Publishing, 2018, pp. 830–840.[35] R. Liu and X. Hu, “A Multimodal Music Recommendation Systemwith Listeners’ Personality and Physiological Signals,” in
Proceedingsof the ACM/IEEE Joint Conference on Digital Libraries in 2020 .New York, NY, USA: ACM, aug 2020, pp. 357–360. [Online].Available: https://dl.acm.org/doi/10.1145/3383583.3398623[36] M. Tkalcic, M. Kunaver, J. Tasic, and A. Koˇsir, “Personality baseduser similarity measure for a collaborative recommender system,” in
Proceedings of the 5th Workshop on Emotion in Human-ComputerInteraction-Real world challenges . 30–37, 2009, pp. 30–37.[37] A. Felfernig, C. Zehentner, G. Ninaus, H. Grabner, W. Maalej,D. Pagano, L. Weninger, and F. Reinfrank, “Group decision supportfor requirements negotiation,” in
International Conference on UserModeling, Adaptation, and Personalization . Springer, 2011, pp. 105–116.[38] I. Cantador and I. Fern´andez-Tob \ ’ \ ias, “On the exploitation of userpersonality in recommender systems,” in CEUR Workshop Proceedings .Mouzhi Ge, 2014.[39] N. Neehal and M. A. Mottalib, “Prediction of Preferred Personality forFriend Recommendation in Social Networks using Artificial NeuralNetwork,” in . IEEE, feb 2019, pp. 1–6.[Online]. Available: https://ieeexplore.ieee.org/document/8679375/[40] O. Nalmpantis and C. Tjortjis, “The 50/50 Recommender: A MethodIncorporating Personality into Movie Recommender Systems,” in
En-gineering Applications of Neural Networks , G. Boracchi, L. Iliadis,C. Jayne, and A. Likas, Eds. Cham: Springer International Publishing,2017, pp. 498–507.[41] W. Wu, L. Chen, and Y. Zhao, “Personalizing recommendationdiversity based on user personality,”
User Modeling and User-AdaptedInteraction , vol. 28, no. 3, pp. 237–276, aug 2018. [Online]. Available:http://link.springer.com/10.1007/s11257-018-9205-x[42] V. Balakrishnan and H. Arabi, “HyPeRM: A HYBRIDPERSONALITY-AWARE RECOMMENDER FOR MOVIE,”
Malaysian Journal of Computer Science , vol. 31, no. 1, pp.48–62, jan 2018. [Online]. Available: https://ejournal.um.edu.my/index.php/MJCS/article/view/10568 [43] M. Schedl, M. Melenhorst, C. C. S. Liem, A. Martorell, ´O. Mayor,and M. Tkalˇciˇc, “A personality-based adaptive system for visualizingclassical music performances,” in
Proceedings of the 7th InternationalConference on Multimedia Systems - MMSys ’16 . New York,New York, USA: ACM Press, 2016, pp. 1–7. [Online]. Available:http://dl.acm.org/citation.cfm?doid=2910017.2910604[44] T. L. Ting and K. D. Varathan, “JOB RECOMMENDATIONUSING FACEBOOK PERSONALITY SCORES,”
Malaysian Journalof Computer Science , vol. 31, no. 4, pp. 311–331, oct2018. [Online]. Available: https://ejournal.um.edu.my/index.php/MJCS/article/view/14244[45] G. Sofia, S. Marianna, L. George, and K. Panos, “Investigating the roleof personality traits and influence strategies on the persuasive effectof personalized recommendations,” in , 2016, p. 9.[46] B. Ferwerda, M. P. Graus, A. Vall, M. Tkalcic, and M. Schedl, “Theinfluence of users’ personality traits on satisfaction and attractivenessof diversified recommendation lists,” in
Empire RecSys . CEUR-WS,2016.[47] Z. Yusefi Hafshejani, M. Kaedi, and A. Fatemi, “Improvingsparsity and new user problems in collaborative filtering byclustering the personality factors,”
Electronic Commerce Research ,vol. 18, no. 4, pp. 813–836, dec 2018. [Online]. Available:http://link.springer.com/10.1007/s10660-018-9287-x[48] M. J. Shayegan and M. Valizadeh, “A Recommender System basedon the analysis of personality traits in Telegram social network,” oct2020. [Online]. Available: http://arxiv.org/abs/2010.00643[49] R. Subramanian, J. Wache, M. K. Abadi, R. L. Vieriu, S. Winkler, andN. Sebe, “ASCERTAIN: Emotion and Personality Recognition UsingCommercial Sensors,”
IEEE Transactions on Affective Computing ,vol. 9, no. 2, pp. 147–160, apr 2018. [Online]. Available:https://ieeexplore.ieee.org/document/7736040/[50] B. Lepri, R. Subramanian, K. Kalimeri, J. Staiano, F. Pianesi,and N. Sebe, “Connecting Meeting Behavior with Extraversion—A Systematic Study,”
IEEE Transactions on Affective Computing ,vol. 3, no. 4, pp. 443–455, 2012. [Online]. Available: http://ieeexplore.ieee.org/document/6212435/[51] G. Zen, B. Lepri, E. Ricci, and O. Lanz, “Space speaks: towardssocially and personality aware visual surveillance,” in
Proceedings ofthe 1st ACM international workshop on Multimodal pervasive videoanalysis , 2010, pp. 37–42.[52] R. Subramanian, Y. Yan, J. Staiano, O. Lanz, and N. Sebe, “Onthe relationship between head pose, social attention and personalityprediction for unstructured and dynamic group interactions,” in
Proceedings of the 15th ACM on International conference onmultimodal interaction - ICMI ’13 . New York, New York,USA: ACM Press, 2013, pp. 3–10. [Online]. Available: http://dl.acm.org/citation.cfm?doid=2522848.2522862[53] M. Perugini and L. Di Blas, “Analyzing personality related adjectivesfrom an eticemic perspective: the big five marker scales (BFMS) andthe Italian AB5C taxonomy,”
Big Five Assessment , pp. 281–304, 2002.[54] G. Roffo, “Towards personality-aware recommendation,” arXiv preprintarXiv:1607.05088 , 2016.[55] S. C. Guntuku, S. Roy, and L. Weisi, “Personality Modeling BasedImage Recommendation,” in
MultiMedia Modeling , X. He, S. Luo,D. Tao, C. Xu, J. Yang, and M. A. Hasan, Eds. Cham: SpringerInternational Publishing, 2015, pp. 171–182.[56] G. Roffo and A. Vinciarelli, “Personality in Computational Advertising:A Benchmark,” 2016. [Online]. Available: http://eprints.gla.ac.uk/149660/[57] H. Ning, S. Dhelim, and N. Aung, “PersoNet: Friend RecommendationSystem Based on Big-Five Personality Traits and Hybrid Filtering,”
IEEE Transactions on Computational Social Systems , pp. 1–9, 2019.[Online]. Available: https://ieeexplore.ieee.org/document/8675299/[58] M. S. H. Mukta, M. E. Ali, and J. Mahmud, “Identifying andvalidating personality traits-based homophilies for an egocentricnetwork,”
Social Network Analysis and Mining , vol. 6, no. 1, p. 74,2016. [Online]. Available: https://doi.org/10.1007/s13278-016-0383-4[59] I. Fern´andez-Tob´ıas, M. Braunhofer, M. Elahi, F. Ricci, andI. Cantador, “Alleviating the new user problem in collaborative filteringby exploiting personality information,”
User Modeling and User-Adapted Interaction , vol. 26, no. 2-3, pp. 221–255, jun 2016. [Online].Available: http://link.springer.com/10.1007/s11257-016-9172-z[60] W. Youyou, D. Stillwell, H. A. Schwartz, and M. Kosinski,“Birds of a Feather Do Flock Together,”
Psychological Science ,vol. 28, no. 3, pp. 276–284, mar 2017. [Online]. Avail-
UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 14 able: http://journals.sagepub.com/doi/10.1177/0956797617697667http://journals.sagepub.com/doi/10.1177/0956797616678187[61] M. Braunhofer, M. Elahi, and F. Ricci, “User Personality and the NewUser Problem in a Context-Aware Point of Interest RecommenderSystem,” in
Information and Communication Technologies inTourism 2015 . Cham: Springer International Publishing, 2015,pp. 537–549. [Online]. Available: http://link.springer.com/10.1007/978-3-319-14343-9 { }
Journal of Research in Personality , vol. 43, no. 3,pp. 524–527, jun 2009. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/S0092656609000439[63] Y. R. Tausczik and J. W. Pennebaker, “The Psychological Meaningof Words: LIWC and Computerized Text Analysis Methods,”
Journalof Language and Social Psychology , vol. 29, no. 1, pp. 24–54, mar2010. [Online]. Available: http://journals.sagepub.com/doi/10.1177/0261927X09351676[64] S. Han, H. Huang, and Y. Tang, “Knowledge of words: An interpretableapproach for personality recognition from social media,”
Knowledge-Based Systems , vol. 194, p. 105550, apr 2020. [Online]. Available:https://linkinghub.elsevier.com/retrieve/pii/S0950705120300459[65] L. Li, H. Zhu, S. Zhao, G. Ding, and W. Lin, “Personality-AssistedMulti-Task Learning for Generic and Personalized Image AestheticsAssessment,”
IEEE Transactions on Image Processing , vol. 29,pp. 3898–3910, 2020. [Online]. Available: https://ieeexplore.ieee.org/document/8970458/[66] N. Annalyn, M. W. Bos, L. Sigal, and B. Li, “Predicting Personalityfrom Book Preferences with User-Generated Content Labels,”
IEEETransactions on Affective Computing , pp. 1–1, 2018. [Online].Available: http://ieeexplore.ieee.org/document/8301566/[67] B. Silva and I. Paraboni, “Learning Personality Traits fromFacebook Text,”
IEEE Latin America Transactions , vol. 16,no. 4, pp. 1256–1262, apr 2018. [Online]. Available: https://ieeexplore.ieee.org/document/8362165/[68] X. Sun, B. Liu, Q. Meng, J. Cao, J. Luo, and H. Yin, “Group-levelpersonality detection based on text generated networks,”
World WideWeb , vol. 23, no. 3, pp. 1887–1906, may 2020. [Online]. Available:http://link.springer.com/10.1007/s11280-019-00729-2[69] A. Darliansyah, M. A. Naeem, F. Mirza, and R. Pears, “SENTIPEDE:A Smart System for Sentiment-based Personality Detection from ShortTexts,”
Journal of Universal Computer Science , vol. 25, no. 10, pp.1323–1352, 2019.[70] W. R. dos Santos, R. M. S. Ramos, and I. Paraboni, “Computationalpersonality recognition from Facebook text: psycholinguistic features,words and facets,”
New Review of Hypermedia and Multimedia
Information Processing & Management ,vol. 54, no. 6, pp. 1101–1114, nov 2018. [Online]. Available:https://linkinghub.elsevier.com/retrieve/pii/S0306457317309081[72] C. Segalin, A. Perina, M. Cristani, and A. Vinciarelli, “The PicturesWe Like Are Our Image: Continuous Mapping of Favorite Pictures intoSelf-Assessed and Attributed Personality Traits,”
IEEE Transactionson Affective Computing , vol. 8, no. 2, pp. 268–285, apr 2017.[Online]. Available: http://ieeexplore.ieee.org/document/7378902/[73] B. Ferwerda, M. Schedl, and M. Tkalcic, “Predicting Personality Traitswith Instagram Pictures,” in
Proceedings of the 3rd Workshop onEmotions and Personality in Personalized Systems 2015 - EMPIRE ’15 .New York, New York, USA: ACM Press, 2015, pp. 7–10. [Online].Available: http://dl.acm.org/citation.cfm?doid=2809643.2809644[74] H. Zhu, L. Li, H. Jiang, and A. Tan, “Inferring PersonalityTraits from Attentive Regions of User Liked Images Via WeaklySupervised Dual Convolutional Network,”
Neural Processing Letters ,vol. 51, no. 3, pp. 2105–2121, jun 2020. [Online]. Available:http://link.springer.com/10.1007/s11063-019-09987-7[75] S. C. Guntuku, J. T. Zhou, S. Roy, W. Lin, and I. W. Tsang,“‘Who Likes What and, Why?’ Insights into Modeling Users’Personality Based on Image ‘Likes’,”
IEEE Transactions on AffectiveComputing , vol. 9, no. 1, pp. 130–143, jan 2018. [Online]. Available:http://ieeexplore.ieee.org/document/7491295/[76] L. Zhang, S. Peng, and S. Winkler, “PersEmoN: A Deep Networkfor Joint Analysis of Apparent Personality, Emotion and TheirRelationship,”
IEEE Transactions on Affective Computing , pp. 1–1, 2020. [Online]. Available: https://ieeexplore.ieee.org/document/8897617/ [77] M. M. Tadesse, H. Lin, B. Xu, and L. Yang, “Personality PredictionsBased on User Behavior on the Facebook Social Media Platform,”
IEEE Access , vol. 6, pp. 61 959–61 969, 2018. [Online]. Available:https://ieeexplore.ieee.org/document/8494744/[78] K. Fong and R. A. Mar, “What Does My Avatar Say About Me?Inferring Personality From Avatars,”
Personality and Social PsychologyBulletin , vol. 41, no. 2, pp. 237–249, feb 2015. [Online]. Available:http://journals.sagepub.com/doi/10.1177/0146167214562761[79] G. Nave, J. Minxha, D. M. Greenberg, M. Kosinski, D. Stillwell,and J. Rentfrow, “Musical Preferences Predict Personality: EvidenceFrom Active Listening and Facebook Likes,”
Psychological Science ,vol. 29, no. 7, pp. 1145–1158, jul 2018. [Online]. Available:http://journals.sagepub.com/doi/10.1177/0956797618761659[80] W. Wu and L. Chen, “Implicit Acquisition of User Personality forAugmenting Movie Recommendations,” in
Lecture Notes in ComputerScience . Springer International Publishing, 2015, pp. 302–314.[Online]. Available: https://doi.org/10.1007/978-3-319-20267-9 { }
IEEE Transactions on Computational SocialSystems , pp. 1–13, 2020. [Online]. Available: https://ieeexplore.ieee.org/document/9269396/[82] N. Y. Asabere and A. Acakpovi, “ROPPSA : TV ProgramRecommendation Based on Personality and Social Awareness,”
Mathematical Problems in Engineering
IEEE Systems Journal , vol. 11, no. 4, pp. 2255–2266, dec 2017.[Online]. Available: http://ieeexplore.ieee.org/document/6877610/[84] N. Y. Asabere, A. Acakpovi, and M. B. Michael, “ImprovingSocially-Aware Recommendation Accuracy Through Personality,”
IEEE Transactions on Affective Computing , vol. 9, no. 3, pp. 351–361,jul 2018. [Online]. Available: https://ieeexplore.ieee.org/document/7904698/[85] J. A. Recio-Garcia, G. Jimenez-Diaz, A. A. Sanchez-Ruiz, andB. Diaz-Agudo, “Personality aware recommendations to groups,” in
Proceedings of the third ACM conference on Recommender systems -RecSys ’09 . New York, New York, USA: ACM Press, 2009, p. 325.[Online]. Available: http://portal.acm.org/citation.cfm?doid=1639714.1639779[86] H.-C. Yang and Z.-R. Huang, “Mining personality traits fromsocial messages for game recommender systems,”
Knowledge-BasedSystems , vol. 165, pp. 157–168, feb 2019. [Online]. Available:https://linkinghub.elsevier.com/retrieve/pii/S095070511830577X[87] M. Onori, A. Micarelli, and G. Sansonetti, “A Comparative Analysis ofPersonality-Based Music Recommender Systems.” in
Empire RecSys ,2016, pp. 55–59.[88] I. Fern´andez-Tob´ıas and I. Cantador, “Personality-aware collaborativefiltering: an empirical study in multiple domains with facebook data,”in
International Conference on Electronic Commerce and Web Tech-nologies . Springer, 2014, pp. 125–137.[89] N. Chakrabarty, S. Chowdhury, S. D. Kanni, and S. Mukherjee,“FAFinder: Friend Suggestion System for Social Networking,” in
Intelligent Data Communication Technologies and Internet of Things ,D. J. Hemanth, S. Shakya, and Z. Baig, Eds. Cham: SpringerInternational Publishing, 2020, pp. 51–58.[90] R. Mugge, P. C. Govers, and J. P. Schoormans, “The developmentand testing of a product personality scale,”
Design Studies ,vol. 30, no. 3, pp. 287–302, may 2009. [Online]. Available:https://linkinghub.elsevier.com/retrieve/pii/S0142694X08000859[91] P. Xiao, Y. Fan, and Y. Du, “A personality-aware followee recom-mendation model based on text semantics and sentiment analysis,”in
National CCF Conference on Natural Language Processing andChinese Computing . Springer, 2017, pp. 503–514.[92] S. Dhelim, N. Huansheng, S. Cui, M. Jianhua, R. Huang,and K. I.-K. Wang, “Cyberentity and its consistency in thecyber-physical-social-thinking hyperspace,”
Computers & ElectricalEngineering , vol. 81, p. 106506, jan 2020. [Online]. Available:https://linkinghub.elsevier.com/retrieve/pii/S0045790618334839[93] J.-L. Zhou, Y. Fu, H. Lu, and C.-J. Sun, “From Popularity to Personality— A Heuristic Music Recommendation Method for Niche Market,”
Journal of Computer Science and Technology , vol. 26, no. 5, p. 816,2011. [Online]. Available: https://doi.org/10.1007/s11390-011-0180-5[94] a. I. Hariadi and D. Nurjanah, “Hybrid attribute and personality basedrecommender system for book recommendation,” in
UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 15
Conference on Data and Software Engineering (ICoDSE) . IEEE,nov 2017, pp. 1–5. [Online]. Available: http://ieeexplore.ieee.org/document/8285874/[95] I. Fern´andez-Tob´ıas and I. Cantador, “On the Use of Cross-DomainUser Preferences and Personality Traits in Collaborative Filter-ing,” in
User Modeling, Adaptation and Personalization , F. Ricci,K. Bontcheva, O. Conlan, and S. Lawless, Eds. Cham: SpringerInternational Publishing, 2015, pp. 343–349.[96] J. Wang, “A Collaborative Filtering Systems based on PersonalityInformation,” in
Proceedings of the 2015 International IndustrialInformatics and Computer Engineering Conference
Proceedings of theBrazilian Symposium on Multimedia and the Web . New York,NY, USA: ACM, nov 2020, pp. 157–164. [Online]. Available:https://dl.acm.org/doi/10.1145/3428658.3430969[98] M. Braunhofer, M. Elahi, and F. Ricci, “Usability Assessment of aContext-Aware and Personality-Based Mobile Recommender System,”in
E-Commerce and Web Technologies , M. Hepp and Y. Hoffner, Eds.Cham: Springer International Publishing, 2014, pp. 77–88.[99] H. Feng and X. Qian, “Recommendation via user’s personality andsocial contextual,” in
Proceedings of the 22nd ACM internationalconference on Conference on information & knowledge management- CIKM ’13 . New York, New York, USA: ACM Press, 2013, pp.1521–1524. [Online]. Available: http://dl.acm.org/citation.cfm?doid=2505515.2507834[100] P. Potash and A. Rumshisky, “Recommender System IncorporatingUser Personality Profile through Analysis of Written Reviews.” in
Empire RecSys , 2016, pp. 60–66.[101] S. Khodabandehlou, S. A. Hashemi Golpayegani, and M. ZivariRahman, “An effective recommender system based on personalitytraits, demographics and behavior of customers in time context,”
DataTechnologies and Applications . IEEE, jan 2012, pp. 714–718. [Online].Available: http://ieeexplore.ieee.org/document/6180983/[103] R. Cheng and B. Tang, “A Music Recommendation System Based onAcoustic Features and User Personalities,” in
Trends and Applicationsin Knowledge Discovery and Data Mining , H. Cao, J. Li, and R. Wang,Eds. Cham: Springer International Publishing, 2016, pp. 203–213.[104] V. Tanasescu, C. B. Jones, G. Colombo, M. J. Chorley, S. M.Allen, and R. M. Whitaker, “The Personality of Venues: Places andthe Five-Factors (’Big Five’) Model of Personality,” in . IEEE, jul 2013, pp. 76–81. [Online]. Available:http://ieeexplore.ieee.org/document/6602043/[105] S. Dhelim, H. Ning, and N. Aung, “ComPath: User Interest Miningin Heterogeneous Signed Social Networks for Internet of People,”
IEEE Internet of Things Journal , pp. 1–1, 2020. [Online]. Available:https://ieeexplore.ieee.org/document/9253614/[106] V. Moscato, A. Picariello, and G. Sperli, “An emotional recommendersystem for music,”
IEEE Intelligent Systems , pp. 1–1, 2020. [Online].Available: https://ieeexplore.ieee.org/document/9204829/[107] N. Majumder, S. Poria, A. Gelbukh, and E. Cambria, “DeepLearning-Based Document Modeling for Personality Detection fromText,”
IEEE Intelligent Systems , vol. 32, no. 2, pp. 74–79, mar 2017.[Online]. Available: http://ieeexplore.ieee.org/document/7887639/[108] T. Mikolov, K. Chen, G. Corrado, and J. Dean, “Efficient estimation ofword representations in vector space,” in , 2013.[109] X.-S. Wei, C.-L. Zhang, H. Zhang, and J. Wu, “Deep bimodal regres-sion of apparent personality traits from short video sequences,”
IEEETransactions on Affective Computing , vol. 9, no. 3, pp. 303–315, 2017.[110] C.-S. Jeong, J.-Y. Lee, and K.-D. Jung, “Adaptive RecommendationSystem for Tourism by Personality Type Using Deep Learning,”
International Journal of Internet, Broadcasting and Communication ,vol. 12, no. 1, pp. 55–60, 2020.[111] Z.-W. He, L. Zhang, and F.-Y. Liu, “DiscoStyle: Multi-levelLogistic Ranking for Personalized Image Style Preference Inference,”
International Journal of Automation and Computing , vol. 17, no. 5,pp. 637–651, oct 2020. [Online]. Available: http://link.springer.com/10.1007/s11633-020-1244-1[112] L. Li, H. Zhu, S. Zhao, G. Ding, H. Jiang, and A. Tan, “PersonalityDriven Multi-task Learning for Image Aesthetic Assessment,” in . IEEE, jul 2019, pp. 430–435. [Online]. Available:https://ieeexplore.ieee.org/document/8784759/[113] Z. Zhang, X. Zhao, G. Wang, and X. Bi, “A New Point-of-InterestClassification Model with an Extreme Learning Machine,”
CognitiveComputation , vol. 10, no. 6, pp. 951–964, 2018. [Online]. Available:https://doi.org/10.1007/s12559-018-9599-0[114] A. Tommasel, A. Corbellini, D. Godoy, and S. Schiaffino, “Personality-aware followee recommendation algorithms: An empirical analysis,”
Engineering Applications of Artificial Intelligence , vol. 51, pp. 24–36,may 2016. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/S0952197616000208[115] A. Tommasel, A. Corbellini, D. L. Godoy, and S. Schiaffino, “On theRole of Personality Traits in Followee Recommendation Algorithms,”in
Argentine Symposium on Artificial Intelligence , 2015, pp. 105–112.[Online]. Available: http://sedici.unlp.edu.ar/handle/10915/52104[116] A. Tommasel, A. Corbellini, D. Godoy, and S. Schiaffino, “Exploringthe role of personality traits in followee recommendation,”
OnlineInformation Review . IEEE, nov 2011, pp. 239–244. [Online].Available: http://ieeexplore.ieee.org/document/6103334/[118] A. E. Bolock, A. E. Kady, C. Herbert, and S. Abdennadher, “Towardsa Character-based Meta Recommender for Movies,” in
ComputationalScience and Technology , R. Alfred, Y. Lim, H. Haviluddin, and C. K.On, Eds. Singapore: Springer Singapore, 2020, pp. 627–638.[119] J. Golbeck and E. Norris, “Personality, movie preferences,and recommendations,” in
Proceedings of the 2013 IEEE/ACMInternational Conference on Advances in Social NetworksAnalysis and Mining - ASONAM ’13 . New York, New York,USA: ACM Press, 2013, pp. 1414–1415. [Online]. Available:http://dl.acm.org/citation.cfm?doid=2492517.2492572[120] M. J. Scott, S. C. Guntuku, W. Lin, and G. Ghinea, “Do Personalityand Culture Influence Perceived Video Quality and Enjoyment?”
IEEETransactions on Multimedia , vol. 18, no. 9, pp. 1796–1807, sep 2016.[Online]. Available: http://ieeexplore.ieee.org/document/7480836/[121] M. Kle´c, “The influence of listener personality on music choices,”
Computer Science , vol. 18, 2017.[122] B. Ferwerda, M. Tkalcic, and M. Schedl, “Personality Traits and MusicGenre Preferences: How Music Taste Varies Over Age Groups,” in
Pro-ceedings of the 1st Workshop on Temporal Reasoning in RecommenderSystems (RecTemp) at the 11th ACM Conference on RecommenderSystems, Como, August 31, 2017. , 2017.[123] B. Ferwerda and M. Schedl, “Personality-based user modeling for mu-sic recommender systems,” in
Joint European Conference on MachineLearning and Knowledge Discovery in Databases . Springer, 2016,pp. 254–257.[124] ——, “Enhancing Music Recommender Systems with PersonalityInformation and Emotional States: A Proposal.” in
Umap workshops ,2014.[125] S. C. Guntuku, W. Lin, M. J. Scott, and G. Ghinea, “Modellingthe influence of personality and culture on affect and enjoyment inmultimedia,” in . IEEE, sep 2015, pp. 236–242.[Online]. Available: http://ieeexplore.ieee.org/document/7344577/[126] L. Wu, L. Chen, R. Hong, Y. Fu, X. Xie, and M. Wang, “A HierarchicalAttention Model for Social Contextual Image Recommendation,”
IEEETransactions on Knowledge and Data Engineering , pp. 1–1, 2019.[Online]. Available: https://ieeexplore.ieee.org/document/8700213/[127] F. Gelli, X. He, T. Chen, and T.-S. Chua, “How Personality Affects ourLikes,” in
Proceedings of the 2017 ACM on Multimedia Conference- MM ’17 . New York, New York, USA: ACM Press, 2017, pp.1828–1837. [Online]. Available: http://dl.acm.org/citation.cfm?doid=3123266.3127909[128] J. H. Kim and Y. Kim, “Instagram user characteristics andthe color of their photos: Colorfulness, color diversity, andcolor harmony,”
Information Processing & Management , vol. 56,
UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 16 no. 4, pp. 1494–1505, jul 2019. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/S0306457318304394[129] M. F. Uddin, S. Banerjee, and J. Lee, “Recommender SystemFramework for Academic Choices: Personality Based RecommendationEngine (PBRE),” in . IEEE, jul 2016, pp. 476–483.[Online]. Available: http://ieeexplore.ieee.org/document/7785779/[130] Y. Huang, H. Liu, W. Li, Z. Wang, X. Hu, and W. Wang,“Lifestyles in Amazon: Evidence from online reviews enhancedrecommender system,”
International Journal of Market Research ,vol. 62, no. 6, pp. 689–706, 2020. [Online]. Available: https://doi.org/10.1177/1470785319844146[131] P. Adamopoulos and V. Todri, “Personality-based recommendations:Evidence from amazon. com.” in
RecSys Posters . Springer, 2015.[132] E. S. de Lima, B. Feij´o, and A. L. Furtado, “Player behavior andpersonality modeling for interactive storytelling in games,”
Entertain-ment Computing , vol. 28, pp. 32–48, dec 2018. [Online]. Available:https://linkinghub.elsevier.com/retrieve/pii/S1875952118300120[133] G. Chan, A. Arya, and A. Whitehead, “Keeping Players Engagedin Exergames,” in
Extended Abstracts of the 2018 CHI Conferenceon Human Factors in Computing Systems . New York, NY,USA: ACM, apr 2018, pp. 1–6. [Online]. Available: https://dl.acm.org/doi/10.1145/3170427.3188455[134] V. Zeigler-Hill and S. Monica, “The HEXACO model of personalityand video game preferences,”
Entertainment Computing , vol. 11, pp.21–26, nov 2015. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/S1875952115000105[135] A. Z. Abbasi, D. H. Ting, H. Hlavacs, B. Wilson, U. Rehman,and A. Arsalan, “Personality differences between videogame vs.non-videogame consumers using the HEXACO model,”
CurrentPsychology , may 2020. [Online]. Available: https://doi.org/10.1007/s12144-020-00793-2[136] W. Wang, J. Chen, J. Wang, J. Chen, J. Liu, and Z. Gong, “Trust-Enhanced Collaborative Filtering for Personalized Point of InterestsRecommendation,”
IEEE Transactions on Industrial Informatics ,vol. 16, no. 9, pp. 6124–6132, sep 2020. [Online]. Available:https://ieeexplore.ieee.org/document/8930072/[137] M. Braunhofer, M. Elahi, M. Ge, and F. Ricci, “Context Depen-dent Preference Acquisition with Personality-Based Active Learningin Mobile Recommender Systems,” in
Learning and CollaborationTechnologies. Technology-Rich Environments for Learning and Collab-oration , P. Zaphiris and A. Ioannou, Eds. Cham: Springer InternationalPublishing, 2014, pp. 105–116.[138] M. Sertkan, J. Neidhardt, and H. Werthner, “What is the“Personality” of a tourism destination?”
Information Technology& Tourism , vol. 21, no. 1, pp. 105–133, 2019. [Online]. Available:https://doi.org/10.1007/s40558-018-0135-6[139] L. Bian and H. Holtzman, “Online friend recommendation throughpersonality matching and collaborative filtering,”
Proc. of UBICOMM ,pp. 230–235, 2011.[140] R. P. Karumur, T. T. Nguyen, and J. A. Konstan, “Exploring theValue of Personality in Predicting Rating Behaviors,” in
Proceedingsof the 10th ACM Conference on Recommender Systems . NewYork, NY, USA: ACM, sep 2016, pp. 139–142. [Online]. Available:https://dl.acm.org/doi/10.1145/2959100.2959140[141] L. Quijano-Sanchez, J. A. Recio-Garcia, and B. Diaz-Agudo,“Personality and Social Trust in Group Recommendations,” in . IEEE, oct 2010, pp. 121–126. [Online]. Available:http://ieeexplore.ieee.org/document/5671421/[142] A. Odi´c, M. Tkalˇciˇc, J. Tasiˇc, and A. Koˇsir, “Personality and socialcontext: impact on emotion induction from movies,” in
Empire RecSys .CEUR-WS. org, 2013.[143] W. Wu, L. Chen, and L. He, “Using personality to adjust diversity inrecommender systems,” in
Proceedings of the 24th ACM Conferenceon Hypertext and Social Media - HT ’13 . New York, NewYork, USA: ACM Press, 2013, pp. 225–229. [Online]. Available:http://dl.acm.org/citation.cfm?doid=2481492.2481521[144] M. Tkalcic, G. Semeraro, and M. Gemmis, “Personality and emotionsin decision making and recommender systems,” 2014.[145] E. M. Khan, M. S. H. Mukta, M. E. Ali, and J. Mahmud,“Predicting Users’ Movie Preference and Rating Behavior fromPersonality and Values,”
ACM Transactions on Interactive IntelligentSystems , vol. 10, no. 3, pp. 1–25, nov 2020. [Online]. Available:https://dl.acm.org/doi/10.1145/3338244[146] S. Dhelim, H. Ning, M. A. Bouras, and J. Ma, “Cyber-Enabled Human-Centric Smart Home Architecture,” in . IEEE, oct2018, pp. 1880–1886. [Online]. Available: https://ieeexplore.ieee.org/document/8560294/[147] S. Gupta, P. Gulati, S. Bhatia, and R. Madaan, “An AutomaticApproach to Music Recommendations Based on Individual PersonalityTraits,”
SSRN Electronic Journal
Proceedings of the25th Conference on User Modeling, Adaptation and Personalization .New York, NY, USA: ACM, jul 2017, pp. 285–288. [Online].Available: https://dl.acm.org/doi/10.1145/3079628.3079693[149] R. Hu, “Design and user issues in personality-based recommendersystems,” in
Proceedings of the fourth ACM conference onRecommender systems - RecSys ’10 . New York, New York,USA: ACM Press, 2010, p. 357. [Online]. Available: http://portal.acm.org/citation.cfm?doid=1864708.1864790[150] B. Ferwerda, E. Yang, M. Schedl, and M. Tkalcic, “PersonalityTraits Predict Music Taxonomy Preferences,” in
Proceedings ofthe 33rd Annual ACM Conference Extended Abstracts on HumanFactors in Computing Systems - CHI EA ’15 . New York, NewYork, USA: ACM Press, 2015, pp. 2241–2246. [Online]. Available:http://dl.acm.org/citation.cfm?doid=2702613.2732754[151] A. B. Melchiorre and M. Schedl, “Personality Correlates of MusicAudio Preferences for Modelling Music Listeners,” in
Proceedingsof the 28th ACM Conference on User Modeling, Adaptation andPersonalization . New York, NY, USA: ACM, jul 2020, pp. 313–317.[Online]. Available: https://dl.acm.org/doi/10.1145/3340631.3394874[152] J. Bansal, M. B. Flannery, and M. H. Woolhouse, “Influenceof personality on music-genre exclusivity,”
Psychology of Music ,p. 030573562095361, sep 2020. [Online]. Available: http://journals.sagepub.com/doi/10.1177/0305735620953611[153] P. Kouki, J. Schaffer, J. Pujara, J. O’Donovan, and L. Getoor,“Generating and Understanding Personalized Explanations in HybridRecommender Systems,”
ACM Transactions on Interactive IntelligentSystems , vol. 10, no. 4, pp. 1–40, dec 2020. [Online]. Available:https://dl.acm.org/doi/10.1145/3365843[154] M. Millecamp, N. N. Htun, C. Conati, and K. Verbert, “What’s in aUser? Towards Personalising Transparency for Music RecommenderInterfaces,” in
Proceedings of the 28th ACM Conference onUser Modeling, Adaptation and Personalization . New York,NY, USA: ACM, jul 2020, pp. 173–182. [Online]. Available:https://dl.acm.org/doi/10.1145/3340631.3394844[155] R. Hu and P. Pu, “Exploring personality’s effect on users’ ratingbehavior,” in
Proceedings of the extended abstracts of the 32ndannual ACM conference on Human factors in computing systems -CHI EA ’14 . New York, New York, USA: ACM Press, 2014, pp.2599–2604. [Online]. Available: http://dl.acm.org/citation.cfm?doid=2559206.2581317[156] P. Alves, P. Saraiva, J. Carneiro, P. Campos, H. Martins, P. Novais,and G. Marreiros, “Modeling Tourists’ Personality in RecommenderSystems: How Does Personality Influence Preferences for TouristAttractions?” in
Proceedings of the 28th ACM Conference onUser Modeling, Adaptation and Personalization . New York, NY,USA: ACM, jul 2020, pp. 4–13. [Online]. Available: https://dl.acm.org/doi/10.1145/3340631.3394843[157] G. Laban and T. Araujo, “Don’t take it personally: Resistanceto individually targeted recommendations with anthropomorphicrecommender systems,” Jul 2020. [Online]. Available: psyarxiv.com/w4mkv[158] A. Khelloufi, H. Ning, S. Dhelim, T. Qiu, J. Ma, R. Huang, andL. Atzori, “A Social Relationships Based Service RecommendationSystem For SIoT Devices,”
IEEE Internet of Things Journal , pp.1–1, 2020. [Online]. Available: https://ieeexplore.ieee.org/document/9167284/[159] X. Cai, H. Ning, S. Dhelim, R. Zhou, T. Zhang, Y. Xu,and Y. Wan, “Robot and its living space A roadmap forrobot development based on the view of living space,”
DigitalCommunications and Networks , dec 2020. [Online]. Available:https://linkinghub.elsevier.com/retrieve/pii/S2352864820302881[160] Y. Mehta, S. Fatehi, A. Kazameini, C. Stachl, E. Cambria, andS. Eetemadi, “Bottom-up and top-down: Predicting personality withpsycholinguistic and language model features,” in
Proceedings of theInternational Conference of Data Mining . IEEE, 2020.
UBMITTED TO ARTIFICIAL INTELLIGENCE REVIEW. 17 [161] F. Celli and B. Lepri, “Is Big Five Better than MBTI? A PersonalityComputing Challenge Using Twitter Data.” in
Proceedings of the FifthItalian Conference on Computational Linguistics (CLiC-it 2018) , 2018.[162] M. Qamhieh, H. Sammaneh, and M. N. Demaidi, “PCRS: PersonalizedCareer-Path Recommender System for Engineering Students,”
IEEEAccess , vol. 8, pp. 214 039–214 049, 2020. [Online]. Available:https://ieeexplore.ieee.org/document/9268112/[163] H.-C. Yang, C. S. Lin, Z.-R. Huang, and T.-H. Tsai, “TextMining on Player Personality for Game Recommendation,” in
Proceedings of the 4th Multidisciplinary International Social NetworksConference on ZZZ - MISNC ’17 . New York, New York,USA: ACM Press, 2017, pp. 1–6. [Online]. Available: http://dl.acm.org/citation.cfm?doid=3092090.3092132[164] D. J. Stillwell and M. Kosinski, “myPersonality project: Example ofsuccessful utilization of online social networks for large-scale socialresearch,”
American Psychologist , vol. 59, no. 2, pp. 93–104, 2014.[165] F. M. Harper and J. A. Konstan, “The MovieLens Datasets,”
ACMTransactions on Interactive Intelligent Systems , vol. 5, no. 4, pp. 1–19,jan 2016. [Online]. Available: https://dl.acm.org/doi/10.1145/2827872[166] I. Cantador, I. Fern´andez-Tob \ ’ \ ias, and A. Bellog \ ’ \ in, “Relatingpersonality types with user preferences in multiple entertainmentdomains,” in CEUR workshop proceedings . Shlomo Berkovsky, 2013.[167] R. P. Karumur and J. A. Konstan, “Relating Newcomer Personality toSurvival and Activity in Recommender Systems,” in
Proceedings of the2016 Conference on User Modeling Adaptation and Personalization- UMAP ’16 . New York, New York, USA: ACM Press, 2016,pp. 195–205. [Online]. Available: http://dl.acm.org/citation.cfm?doid=2930238.2930246[168] I. Adaji, C. Sharmaine, S. Debrowney, K. Oyibo, and J. Vassileva,“Personality Based Recipe Recommendation Using Recipe NetworkGraphs,” in
Social Computing and Social Media. Technologies and An-alytics , G. Meiselwitz, Ed. Cham: Springer International Publishing,2018, pp. 161–170.[169] S. Yakhchi, A. Beheshti, S. M. Ghafari, and M. Orgun, “Enabling theAnalysis of Personality Aspects in Recommender Systems,” jan 2020.[Online]. Available: http://arxiv.org/abs/2001.04825[170] W. Wu, “Implicit Acquisition of User Personality for AugmentingRecommender Systems,” in
Proceedings of the 22nd InternationalConference on Intelligent User Interfaces Companion - IUI ’17Companion . New York, New York, USA: ACM Press, 2017,pp. 201–204. [Online]. Available: http://dl.acm.org/citation.cfm?doid=3030024.3038287[171] J. Hinds, E. J. Williams, and A. N. Joinson, ““It wouldn’thappen to me”: Privacy concerns and perspectives following theCambridge Analytica scandal,”
International Journal of Human-Computer Studies , vol. 143, p. 102498, nov 2020. [Online]. Available:https://linkinghub.elsevier.com/retrieve/pii/S1071581920301002
Sahraoui Dhelim
Received his B.S. in ComputerScience from the University of Djelfa, Algeria, in2012 and his Master degree in Networking and Dis-tributed Systems from the University of Laghouat,Algeria, in 2014. And PhD in Computer Scienceand Technology from University of Science andTechnology Beijing, China, in 2020. His currentresearch interests include Social Computing, Per-sonality Computing, User Modeling, Interest Min-ing, Recommendation Systems and Intelligent Trans-portation Systems.
Nyothiri Aung
Received Master of Engineering(Information Technology) degree form MandalayTechnological University, Myanmar, 2012. And PhDin Computer Science and Technology from Uni-versity of Science and Technology Beijing, China,2020. She worked as a tutors at the Departmentof Information Technology in Technological Uni-versity of Meiktila, Myanmar (2008-2010). AndSystem Analyst of ACE Data System, Myanmar(2012-2015). Her research interests include SocialComputing, Personality Computing and IntelligentTransportation System.
Mohammed Amine Bouras
Received B.S. andM.S degree from University of Laghouat, Algeria,the school of computer science in 2015. He iscurrently a Ph.D candidate in University of scienceand Technology Beijing, China, school of computerand communication Engineering. He focuses on theconvergence in the Internet of Things and Big Data.His research interest includes Internet of Things,semantic internet of things, Big Data, Data analysis,Cyber Physical Social Thinking (CPST) Spaces.
Huansheng Ning
Received his B.S. degree fromAnhui University in 1996 and his Ph.D. degreefrom Beihang University in 2001. Now, he is aprofessor and vice dean of the School of Computerand Communication Engineering, University of Sci-ence and Technology Beijing, China. His currentresearch focuses on the Internet of Things and gen-eral cyberspace. He is the founder and chair of theCyberspace and Cybermatics International Scienceand Technology Cooperation Base. He has presidedmany research projects including Natural ScienceFoundation of China, National High Technology Research and DevelopmentProgram of China (863 Project). He has published more than 100+ jour-nal/conference papers, and authored 5 books. He serves as an associate editorof IEEE Systems Journal (2013-Now), IEEE Internet of Things Journal (2014-2018), and as steering committee member of IEEE Internet of Things Journal(2016-Now).