MWE Lexicons for NLP: A State-of-the-Art Survey

This dataset presents the results of a survey on lexicons containing multiword expressions. The resource offers an updated perspective on this field of research as compared to earlier surveys. We identified 66 resources, predominantly compiled from 2016 to 2024, both lexica dedicated to multiword expressions and general lexica featuring multiword expressions. We analyze these 66 resources from several perspectives: types of lexica, types of multiword expressions contained, languages covered, linkage to other resources, acquisition method, and representativeness. Further, we provide a more in-depth analysis of how the lemma of the multiword expressions is encoded in these resources and how their meaning is represented. In the context of the current trend of using large language models in language processing, we find strong indications that the linguistic information from such lexica can help enhance their performance in dealing with multiword expressions. The observations on the existing multiword expressions lexica motivate recommendations for some best practices in their development.