Semantically enhanced term frequency based on word embeddings for Arabic information retrieval

Traditional Information Retrieval (IR) models are based on bag-of-words paradigm, where relevance scores are computed based on exact matching of keywords. Although these models have already achieved good performance, it has been shown that most of dissatisfaction cases in relevance are due to term mismatch between queries and documents. In this paper, we introduce novel method to compute term frequency based on semantic similarities using distributed representations of words in a vector space (Word Embeddings). Our main goal is to allow distinct but semantically related terms to match each other and contribute to the relevance scores. Hence, Arabic documents are retrieved beyond the bag-of-words paradigm based on semantic similarities between word vectors. The results on Arabic standard TREC data sets show significant improvement over the baseline bag-of-words models.