Using Semantic Query Expansion and Relevance Feedback for an Effective Tweet Contextualization Process

Bound to 240 characters, tweets are short and ambiguous by nature due to the way they are written, without maintaining formal grammar and proper spelling. These constraints increase the possibility of the tweet’s misunderstanding. Thus, it is essential to know the original context of their realization. This paper falls under the tweet contex-tualization task which aims to produce an informative and coherent paragraph, called a context, from a set of documents in response to topics treated by the tweet, allowing a reader to better understand the tweet. We propose to combine a semantic query expansion approach and Relevance Feedback technique in order to enhance queries (tweets) and documents (returned by the Information Retrieval System), to produce more informative contexts.The effectiveness of our method is proved through an experimental study conducted on the INEX 2014 collection.