Using Multiword Expression lexicons in Natural Language Processing tasks: A survey
Voula Giouli, Verginica Barbu Mititelu, Raquel Amaro, Svetla Koeva, Gražina Korvel, Irina Lobzhanidze, Giedrė Valūnaitė Oleškevičienė · Zenodo (CERN European Organization for Nuclear Research) · 2026
Multiword expressions remain a persistent challenge for Natural Language Processing tools and applications across a wide range of tasks. This chapter presents and discusses the results of a twofold survey: it combines a comprehensive review of recent literature and a questionnaire for researchers and practitioners in the field. Both show which tasks/applications multiword expression lexicons are used for, what types of information from such resources are relevant for the respective tasks, what types of information could be worth including therein, and to what extent the use of the lexical resource at hand had an impact on the overall performance of the task or of the tool. The survey reveals that MWE lexicons are already used in core NLP tasks like parsing, machine translation, LLM evaluation and others. But in spite of that, richer and standardised annotation schemes and stronger links between lexicons and corpora are still needed to support their effectiveness.