Marcos Aurélio Domingues

Dimensions as Virtual Items: Improving the predictive ability of top-N recommender systems

Traditionally, recommender systems for the web deal with applications that have two dimensions, users and items. Based on access data that relate these dimensions, a recommendation model can be built and used to identify a set of N items that will be of interest to a certain user. In this paper we propose a multidimensional approach, called DaVI (Dimensions as Virtual Items), that consists in inserting contextual and background information as new user-item pairs. The main advantage of this approach is that it can be applied in combination with several existing two-dimensional recommendation algorithms. To evaluate its effectiveness, we used the DaVI approach with two different top-N recommender algorithms, Item-based Collaborative Filtering and Association Rules based, and ran an extensive set of experiments in three different real world data sets. In addition, we have also compared our approach to the previously introduced combined reduction and weight post-filtering approaches. The empirical results strongly indicate that our approach enables the application of existing two-dimensional recommendation algorithms in multidimensional data, exploiting the useful information of these data to improve the predictive ability of top-N recommender systems.

Using statistics, visualization and data mining for monitoring the quality of meta-data in web portals

The goal of many web portals is to select, organize and distribute content in order to satisfy its users/customers. This process is usually based on meta-data that represent and describe content. In this paper we describe a methodology and a system to monitor the quality of the meta-data used to describe content in web portals. The methodology is based on the analysis of the meta-data using statistics, visualization and data mining tools. The methodology enables the site’s editor to detect and correct problems in the description of contents, thus improving the quality of the web portal and the satisfaction of its users. We also define a general architecture for a system to support the proposed methodology. We have implemented this system and tested it on a Portuguese portal for management executives. The results validate the methodology proposed.

Combining usage and content in an online music recommendation system for music in the long-tail

In this paper we propose a hybrid music recommender system, which combines usage and content data. We describe an online evaluation experiment performed in real time on a commercial music web site, specialised in content from the very long tail of music content. We compare it against two stand-alone recommenders, the first system based on usage and the second one based on content data. The results show that the proposed hybrid recommender shows advantages with respect to usage- and content-based systems, namely, higher user absolute acceptance rate, higher user activity rate and higher user loyalty.

Privileged contextual information for context-aware recommender systems

WA method that treats additional information as virtual items in recommender systems.The method is instantiated in three different algorithms for recommender systems.The algorithms are evaluated among themselves and against the state-of-the-art.Results show that the proposal improves the predictive ability of the recommenders. A recommender system is used in various fields to recommend items of interest to the users. Most recommender approaches focus only on the users and items to make the recommendations. However, in many applications, it is also important to incorporate contextual information into the recommendation process. Although the use of contextual information has received great focus in recent years, there is a lack of automatic methods to obtain such information for context-aware recommender systems. Some works address this problem by proposing supervised methods, which require greater human effort and whose results are not so satisfactory. In this scenario, we propose an unsupervised method to extract contextual information from web page content. Our method builds topic hierarchies from page textual content considering, besides the traditional bag-of-words, valuable information of texts as named entities and domain terms (privileged information). The topics extracted from the hierarchies are used as contextual information in context-aware recommender systems. We conducted experiments by using two data sets and two baselines: the first baseline is a recommendation system that does not use contextual information and the second baseline is a method proposed in literature to extract contextual information. The results are, in general, very good and present significant gains. In conclusion, our method has advantages and innovations:(i) it is unsupervised; (ii) it considers the context of the item (Web page), instead of the context of the user as in most of the few existing methods, which is an innovation; (iii) it uses privileged information in addition to the existing technical information from pages; and (iv) it presented good and promising empirical results. This work represents an advance in the state-of-the-art in context extraction, which means an important contribution to context-aware recommender systems, a kind of specialized and intelligent system.

Using Contextual Information from Topic Hierarchies to Improve Context-Aware Recommender Systems

Unlike the traditional recommender systems, that make recommendations only by using the relation between user and item, a context-aware recommender system makes recommendations by incorporating available contextual information into the recommendation process as explicit additional categories of data to improve the recommendation process. In this paper, we propose to use contextual information from topic hierarchies to improve the accuracy of context-aware recommender systems. Additionally, we also propose two context-aware recommender algorithms for item recommendation. These are extensions from algorithms proposed in literature for rating prediction. The empirical results demonstrate that by using topic hierarchies our technique can provide better recommendations.

Improving Personalized Ranking in Recommender Systems with Topic Hierarchies and Implicit Feedback

The knowledge of semantic information about the content and users preferences is an important issue to improve recommender systems. However, the extraction of such meaningful metadata needs an intense and time-consuming human effort, which is impractical specially with large databases. In this paper, we mitigate this problem by proposing a recommendation model based on latent factors and implicit feedback which uses an unsupervised topic hierarchy constructor algorithm to organize and collect metadata at different granularities from unstructured textual content. We provide an empirical evaluation using a dataset of web pages written in Portuguese language, and the results show that personalized ranking with better quality can be generated using the extracted topics at medium granularity.

Applying multi-view based metadata in personalized ranking for recommender systems

In this paper, we propose a multi-view based metadata extraction technique from unstructured textual content in order to be applied in recommendation algorithms based on latent factors. The solution aims at reducing the problem of intense and time-consuming human effort to identify, collect and label descriptions about the items. Our proposal uses a unsupervised learning method to construct topic hierarchies with named entity recognition as privileged information. We evaluate the technique using different recommendation algorithms, and show that better accuracy is obtained when additional information about items is considered.

Exploiting Text Mining Techniques for Contextual Recommendations

Unlike traditional recommender systems, which make recommendations only by using the relation between users and items, a context-aware recommender system makes recommendations by incorporating available contextual information into the recommendation process. One problem of context-aware approaches is that it is required techniques to extract such additional information in an automatic manner. In this paper, we propose to use two text mining techniques which are applied to textual data to infer contextual information automatically: named entities recognition and topic hierarchies. We evaluate the proposed technique in four context-aware recommender systems. The empirical results demonstrate that by using named entities and topic hierarchies we can provide better recommendations.

Post-processing of Association Rules using Taxonomies

The data mining process enables the end users to analyse, understand and use the extracted knowledge in an intelligent system or to support in the decision-making processes. However, many algorithms used in the process encounter large quantities of patterns, complicating the analysis of the patterns. This fact occurs with association rules, a data mining technique that tries to identify intrinsic patterns in large data sets. A method that can help the analysis of the association rules is the use of taxonomies in the step of post-processing knowledge. In this paper, the GART algorithm is proposed, which uses taxonomies to generalize association rules, and the RulEE-GAR computational module, that enables the analysis of the generalized rules

Improving Personalized Ranking in Recommender Systems with Multimodal Interactions

This paper proposes a conceptual framework which uses multimodal user feedback to generate a more accurate personalized ranking of items to the user. Our technique is a response to the actual scenario on the Web, where users can consume content following different interaction paradigms, such as rating, browsing, sharing, etc. We developed a post-processing step to ensemble rankings generated by unimodal-based state-of-art algorithms, using a set of heuristics which analyze the behavior of the user during consumption. We provide an experimental evaluation using the Movie Lens 10M dataset, and the results show that better recommendations can be provided when multimodal interactions are considered for profiling the preferences of the users.