Quality of word and concept embeddings in targetted biomedical domains

Embeddings are fundamental resources often reused for building intelligent systems in thebiomedical context. As a result, evaluating the quality of previously trained embeddings andensuring they cover the desired information is critical for the success of applications. Thispaper proposes a new evaluation methodology to test the coverage of embeddings against atargetted domain of interest. It defines measures to assess the terminology, similarity, and analogycoverage, which are core aspects of the embeddings. Then, it discusses the experimentationcarried out on existing biomedical embeddings in the specific context of pulmonary diseases.The proposed methodology and measures are general and may be applied to any applicationdomain.