ECNU at SemEval-2017 Task 7: Using Supervised and Unsupervised Methods to Detect and Locate English Puns

Proceedings of the 11th International Workshop on Semantic Evaluations (SemEval-2017), pages 453–456, Vancouver, Canada, August 3 - 4, 2017. c⃝2017 Association for Computational Linguistics ECNU at SemEval-2017 Task 7: Using Supervised and Unsupervised Methods to Detect and Locate English Puns Yuhuan Xiu1, Man Lan1,2∗, Yuanbin Wu1,2 1Department of Computer Science and Technology, East China Normal University, Shanghai, P.R.China 2Shanghai Key Laboratory of Multidimensional Information Processing 51164500032@stu.ecnu.edu.cn mlan,ybwu@cs.ecnu.edu.cn Abstract This paper describes our submissions to task 7 in SemEval 2017, i.e., Detection and Interpretation of English Puns. We par- ticipated in the first two subtasks, which are to detect and locate English puns re- spectively. For subtask 1, we presented a supervised system to determine whether or not a sentence contains a pun using simi- larity features calculated on sense vectors or cluster center vectors. For subtask 2, we established an unsupervised system to locate the pun by scoring each word in the sentence and we assumed that the word with the smallest score is the pun. 1 Introduction A pun is a form of wordplay in which one signifier (e.g., a word or phrase) suggests two or more meanings by exploiting polysemy, or phonological similarity to another signifier, for an intended humorous or rhetorical effect. The study of puns can be seen as a respectable research topic in traditional linguistics and the cognitive sciences. Semeval 2017 task 7(Miller et al., 2017) con- tains three subtasks, i.e., pun detection, pun loca- tion, and pun interpretation. And we participated in the first two subtasks. The detection and location of English puns are to determine whether or not a sentence contains a pun and which word is a pun respectively, which differ from traditional word sense disambiguation (WSD). WSD is to determine an exact meaning of the target word in the given context. However, WSD algorithms could provide the lexical-semantic understanding for pun detection and location. And we adopted a knowledge-based WSD algorithm to obtain possi- ble senses1 for each word in the sentence. 1The sense is the gloss provided by WordNet There are two types of puns: some are homo- graphic puns and the others are heterographic pun- s. A homographic pun exploits distinct meanings of the same written word, and a heterographic pun exploits distinct meanings of the similar but not exactly the same spoken word. The organizer- s provided two test datasets about homographic puns and heterographic puns respectively for each subtask. Since they did not provide official train- ing datasets, we collected our own positive sam- ples(each sentence contains a pun) from the Pun of the Day website2, which conclude 60 homo- graphic puns and 60 heterographic puns. Besides, we also assembled a raw dataset of 120 negative samples(sentences that do not contain puns) from the Internet. Then, we combined 120 negative samples with 60 homographic puns or 60 hetero- graphic puns into homographic or heterographic training dataset. Then, we did the same data s.preprocessing for both training and test dataset- Firstly, we performed part-of-speech(POS) tagging using Stanford CoreNLP …

Content truncated to 500 words