Bridging Semantic Gaps in Information Retrieval: Context-Based Approaches

In Information Retrieval (IR), the semantic gap is the difference between what computers store and what users expect via their queries. There are several reasons for the existence of those gaps such as homonymy and synonymy in text retrieval, or the typical difference between low-level representations and keyword-based queries in image retrieval. The objective of this work is to close these gaps by effective, scalable and not-so-expensive solutions. The main idea is to exploit available unstructured data and hidden topic models to infer surrounding contexts for better information retrieval (in both text retrieval and image retrieval). Early results obtained on two problems, namely Web search clustering and image annotation, show the effectiveness of the proposed approaches.