Classifier Combination for Contextual Idiom Detection Without Labelled Data

Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing, pages 315–323, Singapore, 6-7 August 2009. c⃝2009 ACL and AFNLP Classifier Combination for Contextual Idiom Detection Without Labelled Data Linlin Li and Caroline Sporleder Saarland University Postfach 15 11 50 66041 Saarbr¨ucken Germany {linlin,csporled}@coli.uni-saarland.de Abstract We propose a novel unsupervised approach for distinguishing literal and non-literal use of idiomatic expressions. Our model com- bines an unsupervised and a supervised classifier. The former bases its decision on the cohesive structure of the context and labels training data for the latter, which can then take a larger feature space into account. We show that a combination of both classi- fiers leads to significant improvements over using the unsupervised classifier alone. 1 Introduction Idiomatic expressions are abundant in natural lan- guage. They also often behave idiosyncratically and are therefore a significant challenge for natural language processing systems. For example, idioms can violate selectional restrictions (as in push one’s luck), disobey typical subcategorisation constraints (e.g., in line without a determiner before line), or change the default assignments of semantic roles to syntactic categories (e.g., in break sth with X the argument X would typically be an instrument but for the idiom break the ice it is more likely to fill a patient role, as in break the ice with Russia). In order to deal with such idiosyncracies and as- sign the correct analyses, NLP systems need to be able to recognise idiomatic expressions. Much pre- vious research on idioms has been concerned with type-based classification, i.e., dividing expressions into ‘idiom’ or ‘not idiom’ irrespective of their ac- tual use in a given context. However, while some expressions, such as by and large, always have an idiomatic meaning, several other expressions, such as break the ice or spill the beans, can be used liter- ally as well as idiomatically (see examples (1) and (2), respectively). Sometimes the literal usage can even dominate in a domain, as for drop the ball, which occurs fairly frequently in a literal sense in the sports section of news texts. (1) Dad had to break the ice on the chicken troughs so that they could get water. (2) Somehow I always end up spilling the beans all over the floor and looking foolish when the clerk comes to sweep them up. Hence, whether a particular occurrence of a po- tentially ambiguous expression has literal or non- literal meaning has to be inferred from the context (token-based idiom classification). Recently, there has been increasing interest in this classification task and both supervised and unsupervised tech- niques have been proposed. The work we present here builds on previous research by Sporleder and Li (2009), who describe an unsupervised method that exploits the presence or absence of cohesive ties between the component words of a potential idiom and its context to distinguish between literal and non-literal use. If strong ties can be found the expression is classified as literal otherwise as non-literal. While this approach often works fairly well, it has …

Content truncated to 500 words