Verbal Multiword Expression Identification: Do We Need a Sledgehammer to Crack a Nut?

Abstract

Automatic identification of multiword expressions (MWEs), like to cut corners 'to do an incomplete job ', is a pre-requisite for semantically-oriented downstream applications. This task is challenging because MWEs, especially verbal ones (VMWEs), exhibit surface variability. This paper deals with a subproblem of VMWE identification: the identification of occurrences of previously seen VMWEs. A simple language-independent system based on a combination of filters competes with the best systems from a recent shared task: it obtains the best averaged F-score over 11 languages (0.6653) and even the best score for both seen and unseen VMWEs due to the high proportion of seen VMWEs in texts. This highlights the fact that focusing on the identification of seen VMWEs could be a strategy to improve VMWE identification in general.This work is licensed under a Creative Commons Attribution 4.0 International License. License details: http:// creativecommons.org/licenses/by/4.0/.1 Henceforth, the lexicalized components of a MWE, i.e. those always realized by the same lexemes, appear in bold.2 Henceforth, literal and coincidental occurrences are highlighted with wavy underlining, following Savary et al. (2019b).

Assessing linguistic generalisation in language models: a dataset for Brazilian Portuguese

“…Evaluations related to MWEs have often focused on specific tasks (Constant et al, 2017) like, for example, on MWE token identification (Scholivet and Ramisch, 2017;Pasquer et al, 2020), evaluating the sensitivity of a model to the occurrence of MWEs in sentences or looking at their interpretation (Cordeiro et al, 2019;Garcia et al, 2021a), assessing to what extent models are able to represent their potential idiomaticity (Tayyar Madabushi et al, 2021, 2022. In particular the results obtained for MWE interpretation by a variety of off-the-shelf embeddings with different levels of contextualisation (including transformer-based models) indicated that idiomaticity was not yet accurately captured even when adopting large pre-trained language models, for any of the two languages covered, English and Portuguese (Garcia et al, 2021b,a).…”

A Survey of MWE Identification Experiments: The Devil is in the Details

“…Figure 1 shows the distribution of papers across the 24 languages considered by our paper sample. The reasons that lead to choosing a given corpus and/or set of languages in non-ST works are various: language diversity (Zampieri et al, 2019), corpus domain (Liu et al, 2021), and corpus quality and size (Pasquer et al, 2020b).…”

A Phraseology Approach in Developmental Education Placement

“…Even though this study focused on L1 and L2 students, it supports the investigation of "MWE across proficiency levels or novice and professional academic writing [...] in an academic context." [11, p.11] The ongoing variability of MWE, particularly verbal ones, and the challenges it poses for automated machine learning identification tasks are recognized by [13]. The authors focused on strategies to improve the identification of verbal MWE (VMWE) and used the PARSEME 7 corpora as a starting point.…”

“…The authors focused on strategies to improve the identification of verbal MWE (VMWE) and used the PARSEME 7 corpora as a starting point. They completed three tasks comprised of a training and development phase, a prediction phase, and an evaluation phase all aided by a simple [candidate VMWE potential] extraction of filtering (Seen2020) techniques for precision [13]. Promising results were evidenced in boosting the identification of global MWE by using a combination of morphosyntactic filters.…”

“…Suggestions for further work emphasize the representation of "VMWE as multisets of (lemma-POS) pairs rather than lemma and POS multisets separately." [13, p.3342] A framework to further investigate the lexical and syntactical features of MWE is provided by [8,11,13,14], among others. However, it is unclear whether automatic placement classification systems, i.e., accuplacer, compass, and act, take MWE into consideration or at least what types of MWE are considered.…”