Francesco Osborne

Mining semantic relations between research areas

For a number of years now we have seen the emergence of repositories of research data specified using OWL/RDF as representation languages, and conceptualized according to a variety of ontologies. This class of solutions promises both to facilitate the integration of research data with other relevant sources of information and also to support more intelligent forms of querying and exploration. However, an issue which has only been partially addressed is that of generating and characterizing semantically the relations that exist between research areas. This problem has been traditionally addressed by manually creating taxonomies, such as the ACM classification of research topics. However, this manual approach is inadequate for a number of reasons: these taxonomies are very coarse-grained and they do not cater for the fine-grained research topics, which define the level at which typically researchers (and even more so, PhD students) operate. Moreover, they evolve slowly, and therefore they tend not to cover the most recent research trends. In addition, as we move towards a semantic characterization of these relations, there is arguably a need for a more sophisticated characterization than a homogeneous taxonomy, to reflect the different ways in which research areas can be related. In this paper we propose Klink, a new approach to i) automatically generating relations between research areas and ii) populating a bibliographic ontology, which combines both machine learning methods and external knowledge, which is drawn from a number of resources, including Google Scholar and Wikipedia. We have tested a number of alternative algorithms and our evaluation shows that a method relying on both external knowledge and the ability to detect temporal relations between research areas performs best with respect to a manually constructed standard.

Propagating User Interests in Ontology-Based User Model

In this paper we address the problem of propagating user interests in ontology-based user models. Our ontology-based user model (OBUM) is devised as an overlay over the domain ontology. Using ontologies as the basis of the user profile allows the initial user behavior to be matched with existing concepts in the domain ontology. Such ontological approach to user profiling has been proven successful in addressing the cold-start problem in recommender systems, since it allows for propagation from a small number of initial concepts to other related domain concepts by exploiting the ontological structure of the domain. The main contribution of the paper is the novel algorithm for propagation of user interests which takes into account i) the ontological structure of the domain and, in particular, the level at which each domain item is found in the ontology; ii) the type of feedback provided by the user, and iii) the amount of past feedback provided for a certain domain object.

User data distributed on the social web: how to identify users on different social systems and collecting data about them

This paper presents an approach to uniquely identify users and to retrieve their data distributed in profiles stored in different systems. The objective is exploiting the public user data available in the Web and especially in social networks. The approach does not require the implementation of specific protocols and the provision of authentication data. The evaluation provides good results that encourage us in carrying on the extension of the project. The extension we are working on is aimed at aggregating, using heuristic techniques, the data stored in the retrieved profiles and at inferring new data about the user.

Identifying Diachronic Topic-Based Research Communities by Clustering Shared Research Trajectories

Communities of academic authors are usually identified by means of standard community detection algorithms, which exploit ‘static’ relations, such as co-authorship or citation networks. In contrast with these approaches, here we focus on diachronic topic-based communities –i.e., communities of people who appear to work on semantically related topics at the same time. These communities are interesting because their analysis allows us to make sense of the dynamics of the research world –e.g., migration of researchers from one topic to another, new communities being spawn by older ones, communities splitting, merging, ceasing to exist, etc. To this purpose, we are interested in developing clustering methods that are able to handle correctly the dynamic aspects of topic-based community formation, prioritizing the relationship between researchers who appear to follow the same research trajectories. We thus present a novel approach called Temporal Semantic Topic-Based Clustering (TST), which exploits a novel metric for clustering researchers according to their research trajectories, defined as distributions of semantic topics over time. The approach has been evaluated through an empirical study involving 25 experts from the Semantic Web and Human-Computer Interaction areas. The evaluation shows that TST exhibits a performance comparable to the one achieved by human experts.

Automatic Classification of Springer Nature Proceedings with Smart Topic Miner

The process of classifying scholarly outputs is crucial to ensure timely access to knowledge. However, this process is typically carried out manually by expert editors, leading to high costs and slow throughput. In this paper we present Smart Topic Miner (STM), a novel solution which uses semantic web technologies to classify scholarly publications on the basis of a very large automatically generated ontology of research areas. STM was developed to support the Springer Nature Computer Science editorial team in classifying proceedings in the LNCS family. It analyses in real time a set of publications provided by an editor and produces a structured set of topics and a number of Springer Nature Classification tags, which best characterise the given input. In this paper we present the architecture of the system and report on an evaluation study conducted with a team of Springer Nature editors. The results of the evaluation, which showed that STM classifies publications with a high degree of accuracy, are very encouraging and as a result we are currently discussing the required next steps to ensure large-scale deployment within the company.

Escaping the Big Brother: An empirical study on factors influencing identification and information leakage on the Web

This paper presents a study on factors that may increase the risks of personal information leakage, owing to the possibility of connecting user profiles that are not explicitly linked together. First, we introduce a technique for user identification based on cross-site checking and linking of user attributes. Then, we describe the experimental evaluation of the identification technique both in a real setting and on an online sample, showing its accuracy to discover unknown personal data. Finally, we combine the results on the accuracy of identification with the results of a questionnaire completed by the same subjects who performed the test in the real setting. The aim of the study was to discover possible factors that make users vulnerable to this kind of technique. We found that the number of social networks used, their features and especially the amount of profiles abandoned and forgotten by the user are factors that increase the likelihood of identification and the privacy risks.

Property-based interest propagation in ontology-based user model

We present an approach for propagation of user interests in ontology-based user models taking into account the properties declared for the concepts in the ontology. Starting from initial user feedback on an object, we calculate user interest in this particular object and its properties and further propagate user interest to other objects in the ontology, similar or related to the initial object. The similarity and relatedness of objects depends on the number of properties they have in common and their corresponding values. The approach we propose can support finer recommendation modalities, considering the user interest in the objects, as well as in singular properties of objects in the recommendation process. We tested our approach for interest propagation with a real adaptive application and obtained an improvement with respect to IS-A-propagation of interest values.

Property-based Semantic Similarity and Relatedness for Improving Recommendation Accuracy and Diversity

The authors introduce new measures of semantic similarity and relatedness for ontological concepts, based on the properties associated to them. They consider two concepts similar if, for some properties they have in common, they also have the same values assigned to these properties. On the other hand, the authors consider two concepts related if they have the same values assigned to different properties. These measures are used in the propagation of user interest values in ontology-based user models to other similar or related concepts in the domain. The authors tested their algorithm in event recommendation domain and in recipe domain and showed that property-based propagation based on similarity outperforms the standard edge-based propagation. Adding relatedness as a criterion for propagation improves diversity without sacrificing accuracy. In addition, assigning a certain relevance to each property improves the accuracy of recommendation. Finally, the property-based spreading activation is effective for cross-domain recommendation.

Anisotropic propagation of user interests in ontology-based user models

This work contributes to the development of ontology-based user models, devised as overlays over conceptual hierarchies derived from domain ontologies. We tackle the problem of propagation of user interests in such a conceptual hierarchy. In addition to accounting for the hierarchical structure of the domain and the type and amount of feedback provided by the user, the principal contributions introduced in this work are: (i) horizontal propagation which enables propagation among siblings, in addition to vertical propagation among ancestors and descendants; (ii) anisotropic vertical propagation which permits user interests to be propagated differently upward and downward; (iii) context-dependance which introduces the possibility to propagate differently according to various contexts for specific applications; (iv) support for dynamic ontology maintenance, i.e. preserving the user interest values when adding or removing a node from the conceptual hierarchy. Our approach supports finer recommendation modalities and contributes to the resolution of the cold start problem, since it allows for propagation from a small number of initial concepts to other related domain concepts by exploiting the conceptual hierarchy of the domain. A field evaluation confirmed the effectiveness of our approach w.r.t. the traditional vertical propagation.

User data discovery and aggregation: The CS-UDD algorithm

In the social web, people use social systems for sharing content and opinions, for communicating with friends, for tagging, etc. People usually have different accounts and different profiles on all of these systems. Several tools for user data aggregation and people search have been developed and protocols and standards for data portability have been defined. This paper presents an approach and an algorithm, named Cross-System User Data Discovery (CS-UDD), to retrieve and aggregate user data distributed on social websites. It is designed to crawl websites, retrieve profiles that may belong to the searched user, correlate them, aggregate the discovered data and return them to the searcher which may, for example, be an adaptive system. The user attributes retrieved, namely attribute–value pairs, are associated with a certainty factor that expresses the confidence that they are true for the searched user. To test the algorithm, we ran it on two popular social networks, MySpace and Flickr. The evaluation has demonstrated the ability of the CS-UDD algorithm to discover unknown user attributes and has revealed high precision of the discovered attributes.