Related terms extraction from Arabic news corpus using word embedding

Abstract

Different techniques are used in text mining to analyze data, extract knowledge, information and relations. We aim in this work to extract related terms for specific keywords. In the first step, we extract Arabic keywords from news articles titles using the TF-IDF terms weighting measure. In the next step, we extract the related terms, from both titles and main texts, using Word2Vec model as a word embedding technique. In order to evaluate our proposed approach, we compute the precision values of the extracted terms that are present in Wikipedia articles. The experiments results perform better for the extracted terms from the articles main texts than titles and the international news category has the highest precision value.