Analyzing User Modeling on Twitter for Personalized News Recommendations

Abstract

Abstract. How can micro-blogging activities on Twitter be leveraged for user modeling and personalization? In this paper we investigate this question and introduce a framework for user modeling on Twitter which enriches the semantics of Twitter messages (tweets) and identifies topics and entities (e.g. persons, events, products) mentioned in tweets. We analyze how strategies for constructing hashtag-based, entity-based or topic-based user profiles benefit from semantic enrichment and explore the temporal dynamics of those profiles. We further measure and compare the performance of the user modeling strategies in context of a personalized news recommendation system. Our results reveal how semantic enrichment enhances the variety and quality of the generated user profiles. Further, we see how the different user modeling strategies impact personalization and discover that the consideration of temporal profile patterns can improve recommendation quality.

TUMS: Twitter-Based User Modeling Service

“…By specifying temporal constraints, client applications can, for example, retrieve the latest profile of a user or a profile that is only based on Twitter activities which a user performed on weekends. In previous work [13], we revealed, for example, that there are significant differences between user profiles created on the weekends with those created during the week.…”

“…First experiments show that entity-based user profiles generated by TUMS, allow for high precision when recommending news articles [13]. Our ambition is to make this news recommender available as Web application using TUMS for generating user profiles.…”

“…To generate such profiles, we developed a set of user modeling strategies [13] that vary in four design dimensions: (i) the type of profiles created by the strategies, (ii) the weighting scheme, (iii) the data sources exploited to further enrich the Twitter-based profiles and (iv) temporal constraints that are considered when constructing the profiles (see Table 1). The generic model for profiles representing users is specified in Definition 1.…”

“…Given that most tweets on Twitter are related to news [3], Twitter seems to be an interesting source of information and TUMS the corresponding service that can support systems that are interested in current interests of a user. In [13] we report on first results in recommending news articles based on TUMS profiles. The TUMS profile can also be used for generating personalized email, recommending webpages, etc.…”

“…In previous work, we proposed strategies to automatically map Twitter messages to related news articles which allows us to exploit the news articles to enrich the semantics of individual tweets [12]. This enrichment builds the basis for the user modeling strategies [13] that are made available via the TUMS service which we present in this paper.…”