Current Trends and Approaches in Synonyms Extraction: Potential Adaptation to Arabic

Extracting synonyms from dictionaries or corpora is gaining special attention as synonyms play an important role in improving NLP application performance. This paper presents a survey of the different approaches and trends used in automatically extracting the synonyms. These approaches can be divided into four main categories. The first approach is to find the Synonyms using a translation graph. The second approach is to discover new transition pairs such as (Arabic-English) (English-France) then (Arabic-France). The third approach is to construct new WordNets by exploring synonymy graphs, and the fourth approach is to find similar words from corpora using Deep Learning methods, such as word embeddings and recently BERT models. The paper also presents a comparative analysis between these approaches and highlights potential adaptation to generate synonyms automatically in the Arabic language as future work.