Building a Semantic Recommendation Engine for News Feeds based on Emerging Topics from Tweets

The rise of social networks powered by the emergence of Web 2.0 unleashed a massive amount of generated user content. Concurrently with technology enhancements that facilitated its widespread, Web 2.0 became the engine which hastened the appearance of worldwide mass communication techniques. Alongside its advent, textual analysis changed as new user-centered content failed to comply with traditional grammar ruling. In this paper, we approach the problem of topic extraction from Twitter in the context of designing a recommendation engine to best matching user profiles to news feed articles. We propose a strategy to extract the concepts by means of Natural Language Processing and use of the semantic cohesion measurements to leverage the matching process. In order to prove the adequacy of our method, we have conducted a medium-scale evaluation. Our results demonstrate the particularities of the Twitter textual corpora, as well as how it can be used to infer geo-locations for its users.