Characterizing Idioms: Conventionality and Contingency

Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 4024 - 4037 May 22-27, 2022 c⃝2022 Association for Computational Linguistics Characterizing Idioms: Conventionality and Contingency Michaela Socolof1,2, Jackie Chi Kit Cheung1,2,3, Michael Wagner1, Timothy J. O’Donnell1,2,3 McGill University1, Quebec AI Institue, Mila2, Canada CIFAR AI Chair3 michaela.socolof@mail.mcgill.ca, chael@mcgill.ca, jcheung@cs.mcgill.ca, timothy.odonnell@mcgill.ca Abstract Idioms are unlike most phrases in two im- portant ways. First, words in an idiom have non-canonical meanings. Second, the non- canonical meanings of words in an idiom are contingent on the presence of other words in the idiom. Linguistic theories differ on whether these properties depend on one an- other, as well as whether special theoretical machinery is needed to accommodate idioms. We define two measures that correspond to the properties above, and we implement them us- ing BERT (Devlin et al., 2019) and XLNet (Yang et al., 2019). We show that English id- ioms fall at the expected intersection of the two dimensions, but that the dimensions them- selves are not correlated. Our results suggest that special machinery to handle idioms may not be warranted. 1 Introduction Idioms—expressions like rock the boat—bring to- gether two phenomena which are of fundamental interest in understanding language. First, they ex- emplify non-conventional word meaning (Wein- reich, 1969; Nunberg et al., 1994). The words rock and boat in this idiom seem to carry par- ticular meanings—something like destabilize and situation, respectively—which are different from the conventional meanings of these words in other contexts. Second, unlike other kinds of non- conventional word use such as novel metaphor, there is a contingency relationship between words in an idiom (Wood, 1986; Pulman, 1993). It is the specific combination of the words rock and boat that has come to carry the idiomatic meaning. Shake the canoe does not have the same accepted meaning. In the literature, most discussions of idioms make use of prototypical examples such as rock the boat. This obscures an important fact: There is no generally agreed-upon definition of idiom; phrase types such as light verb constructions (e.g., take a walk) and semantically transparent colloca- tions (e.g., now or never) are sometimes included in the class (e.g., Palmer, 1981) and sometimes not (e.g., Cowie, 1981). This lack of homogeneity among idiomatic phrases has been recognized as a challenge in the domain of NLP, with Sag et al. (2002) suggesting that a variety of techniques are needed to deal with different kinds of multi-word expressions. What does seem clear is that pro- totypical cases of idiomatic phrases tend to have higher levels of both non-conventional meaning and contingency between words. This combination of non-conventionality and contingency has led to a number of theories that treat idioms as exceptions to the mechanisms that build phrases compositionally. These theories posit special machinery for handling idioms (e.g., Weinreich, 1969; Bobrow and Bell, 1973; Swin- ney and Cutler, 1979). An early but representa- tive example of this position is Weinreich (1969), who posits the addition of two structures to …

Content truncated to 500 words