UWaterloo at SemEval-2017 Task 7: Locating the Pun Using Syntactic Characteristics and Corpus-based Metrics

Proceedings of the 11th International Workshop on Semantic Evaluations (SemEval-2017), pages 421–425, Vancouver, Canada, August 3 - 4, 2017. c⃝2017 Association for Computational Linguistics UWaterloo at SemEval-2017 Task 7: Locating the Pun Using Syntactic Characteristics and Corpus-based Metrics Olga Vechtomova University of Waterloo Waterloo, ON, Canada ovechtom@uwaterloo.ca Abstract The paper presents a system for locat- ing a pun word. The developed method calculates a score for each word in a pun, using a number of components, in- cluding its Inverse Document Frequency (IDF), Normalized Pointwise Mutual In- formation (NPMI) with other words in the pun text, its position in the text, part-of- speech and some syntactic features. The method achieved the best performance in the Heterographic category and the second best in the Homographic. Further analysis showed that IDF is the most useful charac- teristic, whereas the count of words with which the given word has high NPMI has a negative effect on performance. 1 Introduction The pun is defined as “A joke exploiting the dif- ferent possible meanings of a word or the fact that there are words which sound alike but have dif- ferent meanings” (Oxford University Press, 2017). When a pun is a spoken utterance, two types of puns are commonly distinguished: homophonic puns, which exploit different meanings of the same word, and heterophonic puns, in which one or more words have similar but not identical pro- nunciations to some other word or phrase that is alluded to in the pun. The SemEval Task 7 (Miller et al., 2017) focused on the identification of puns as written texts, rather than spoken ut- terances, and hence distinguished between homo- graphic and heterographic puns. We participated in Subtask 2: Pun location, which required participating systems to identify which word is the pun. Only the cases which con- tain exactly one pun word were given to the partic- ipants in each of the two categories: homographic and heterographic puns. Our approach to identifying the pun word is to rank words in the pun text by a score calculated as the sum of values of eleven features. The feature values are calculated using a combination of cor- pus statistics and rule-based methods. The word with the highest score is considered to be the pun word. The method is described in detail in Sec- tion 2. In developing the word ranking method, we were guided by a number of intuitions, out- lined below. The punchline in a pun or a joke is almost al- ways close to the end, since it is at the end that the reader is expected to uncover the second hidden (non-obvious) meaning of the pun. This intuition is consistent with Ruskin’s Script-based Semantic Theory of humour (Ruskin, 1985). The system therefore only assigns scores to words located in the second half of the pun text. What makes a homographic pun humorous is the simultaneous perception by a reader of two conflicting meanings of the same pun word. The pun author can achieve this by using words that …

Content truncated to 500 words