Fermi at SemEval-2017 Task 7: Detection and Interpretation of Homographic puns in English Language

Proceedings of the 11th International Workshop on Semantic Evaluations (SemEval-2017), pages 457–460, Vancouver, Canada, August 3 - 4, 2017. c⃝2017 Association for Computational Linguistics Fermi at SemEval-2017 Task 7: Detection and Interpretation of Homographic puns in English Language Vijayasaradhi Indurthi SIEL, IIIT Hyderabad vijaya.saradhi@research.iiit.ac.in Oota Subba Reddy SIEL, IIIT Hyderabad oota.subba@students.iiit.ac.in Abstract This paper describes our system for de- tection and interpretation of English puns. We participated in 2 subtasks related to homographic puns achieve comparable re- sults for these tasks. Through the paper we provide detailed description of the ap- proach, as well as the results obtained in the task. Our models achieved an F1-score of 77.65% for Subtask 1 and 52.15% for Sub- task 2. 1 Introduction The pun, also called paronomasia, is a form of word play that suggests two or more meanings, by exploiting multiple meanings of words, or of similar-sounding words, for an intended humorous or rhetorical effect. A pun is also a special form of ambiguity (mostly lexical) that is consciously used to create statements with ambiguous-distinct- meanings. These ambiguities can arise from the intentional use of homophonic, homographic, metonymic, or figurative language. A homo- graphic pun exploits distinct meanings of the same written word, and a homophonic pun exploits dis- tinct meanings of the same spoken word. Examples of homographic puns. • ”I used to be a banker but I lost interest”, • ”Tires are fixed for a flat rate” • ”Getting rid of your boat for another could cause a whole raft of problems” In the first example, the word interest is the pun denoting interest as willingness and also as a fixed charge for borrowing money. In the second exam- ple, the word flat is the pun denoting flat as in a flat tyre and flat as in flat rate. The third example, the word raft is the pun denoting raft as a batch and raft as a type of boat. In the present work, we focus on homographic puns. We present methods to a) identify a pun sen- tence, b) identify the pun word given a pun sen- tence and c) interpret the different word senses of the pun word. The details of the shared task are available at (Tristan Miller and Hempelmann, 2017) 2 Subtask 1 - Pun detection In this task, for each sentence, the system must decide whether it contains a pun or not. While this task is mostly unsupervised, we cast this prob- lem as a supervised learning classification prob- lem. We have randomly selected part of the dataset and manually annotated them as a pun sentence or a non-pun. We decided to leverage on models that try to model sequences of word vectors. We can view the each sentence as a sequence where we only have one label at the end. This many-to-one mapping lends itself nicely to Recurrent Neural Network. Due to this reason, we used a recurrent neural network to train the classifier and generate the model. Using this model we classify the …

Content truncated to 500 words