According to tastes, a person could show preference for a given category of content to a greater or lesser extent. However, quantifying people's amount of interest in a certain topic is a challenging task, especially considering the massive digital information they are exposed to. For example, in the context of Twitter, aligned with his/her preferences a user may tweet and retweet more about technology than sports and do not share any music-related content. The problem we address in this paper is the identification of users' implicit topic preferences by analyzing the content categories they tend to post on Twitter. Our proposal is significant given that modeling their multi-topic profile may be useful to find patterns or association between preferences for categories, discover trending topics and cluster similar users to generate better group recommendations of content. In the present work, we propose a method based on the Mixed Gaussian Model to extract the multidimensional preference representation for 399 Ecuadorian tweeters concerning twentytwo different topics (or dimensions) which became known by manually categorizing 68.186 tweets. Our experiment findings indicate that the proposed approach is effective at detecting the topic interests of users.
The rise of a trending topic on Twitter or Facebook leads to the temporal emergence of a set of users currently interested in that topic. Given the temporary nature of the links between these users, being able to dynamically identify communities of users related to this trending topic would allow for a rapid spread of information. Indeed, individual users inside a community might receive recommendations of content generated by the other users, or the community as a whole could receive group recommendations, with new content related to that trending topic. In this paper, we tackle this challenge, by identifying coherent topic-dependent user groups, linking those who generate the content (creators) and those who spread this content, e.g., by retweeting/reposting it (distributors). This is a novel problem on group-to-group interactions in the context of recommender systems. Analysis on real-world Twitter data compare our proposal with a baseline approach that considers the retweeting activity, and validate it with standard metrics. Results show the effectiveness of our approach to identify communities interested in a topic where each includes content creators and content distributors, facilitating users' interactions and the spread of new information.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.