Semantically Enhanced Term Frequency

In this paper, we complement the term frequency, which is used in many bag-of-words based information retrieval models, with information about the semantic relatedness of query and document terms. Our experiments show that when employed in the standard probabilistic retrieval model BM25, the additional semantic information significantly outperforms the standard term frequency, and also improves the effectiveness when additional query expansion is applied. We further analyze the impact of different lexical semantic resources on the IR effectiveness. Key words: Information Retrieval, Semantic Relatedness

• Type: Conference
• Year: 2010
• Where: ECIR
• Authors: Christof Müller, Iryna Gurevych
• Keywords: ECIR 2010, Information Technology, Retrieval Model, Semantic Relatedness, Term Frequency