A Survey of Word Embeddings Evaluation Methods
It provides a systematic overview for researchers and practitioners working with word embeddings, but it is incremental as it compiles existing methods without introducing new ones.
This paper surveys evaluation methods for word embeddings, summarizing 16 intrinsic and 12 extrinsic approaches to address the open problem of determining the most adequate evaluation techniques in the field.
Word embeddings are real-valued word representations able to capture lexical semantics and trained on natural language corpora. Models proposing these representations have gained popularity in the recent years, but the issue of the most adequate evaluation method still remains open. This paper presents an extensive overview of the field of word embeddings evaluation, highlighting main problems and proposing a typology of approaches to evaluation, summarizing 16 intrinsic methods and 12 extrinsic methods. I describe both widely-used and experimental methods, systematize information about evaluation datasets and discuss some key challenges.