Computational phraseology discovery in corpora with the MWETOOLKIT

Computer tools can help discovering new phraseological units in corpora, thanks to their ability to quickly draw statistics from large amounts of textual data. While the research community has focused on developing and evaluating original algorithms for the automatic discovery of phraseological units, little has been done to transform these sophisticated methods into usable software. In this chapter, we present a brief survey of the main approaches to computational phraseology available. Furthermore, we provide worked out examples of how to apply these methods using the mwetoolkit, a free software for the discovery and identification of multiword expressions. The usefulness of the automatically extracted units depends on various factors such as language, corpus size, target units, and available taggers and parsers. Nonetheless, the mwetoolkit allows fine-grained tuning so that this variability is taken into account, adapting the tool to the specificities of each lexicographic environmen