TrentoTeam at SemEval-2017 Task 3: An application of Grice Maxims in Ranking Community Question Answers

Proceedings of the 11th International Workshop on Semantic Evaluations (SemEval-2017), pages 271–274, Vancouver, Canada, August 3 - 4, 2017. c⃝2017 Association for Computational Linguistics TrentoTeam at SemEval-2017 Task 3: An application of Grice Maxims in Ranking Community Question Answers Mohammed R. H. Qwaider Fondazione Bruno Kessler Via Sommarive, 18 Trento, Italy. qwaider@fbk.eu Abed Alhakim Freihat University of Trento Via Sommarive, 19 Trento, Italy. abed.freihat@unitn.it Fausto Giunchiglia University of Trento Via Sommarive, 19 Trento, Italy. fausto@unitn.it Abstract In this paper, we present a community an- swers ranking system which is based on Grice Maxims. In particular, we describe a ranking system which is based on an- swer relevancy scores, assigned by three main components: Named entity recogni- tion, similarity score, and sentiment anal- ysis. 1 Introduction This paper describes the Grice Ranking sys- tem which participated to SemEval 2017 Task 3 (Nakov et al., 2017) subtask A competition1. The SemEval 2017 Task 3 (Nakov et al., 2017) sub-task A focuses on raking a list of answers (10 in our case) as follows. Given a question Q and a list of answers ⟨a1, ..., an⟩. Rank these an- swers according to their relevancy with respect to the question Q. Our participation to the task was mainly motivated by our interest in applying Grice maxims (Grice, 1975) principles to a ranking task to define standards of Grice maxims for ranking tasks. The system follows 3 steps: similarity, en- tity recognition, and sentiment analysis. 2 Grice maxims principles Grice main idea is that communication between human beings is logic and rational. Following this idea, any conversation assumes cooperation be- tween the conversation parties. This cooperation supposes in essence four maxims that usually hold in dialogues or conversations. These maxims are: 1. Quality: Say only true things. 2. Quantity: Be informative. 3. Relation: Be relevant in your conversation. 1http://alt.qcri.org/semeval2016/task3/ 4. Manner: Be direct and straightforward. These maxims have been intensively researched in the domain of linguistics and pragmatics in the last decades, where the researchers focused on how to use Grice theory to explain speaker inten- tion when he says some thing. For example, these maxims explain that the speaker B understands the intention of the speaker A. The same holds for A who understands the indirect Answer of B. A What is the time? B The bus left five minutes ago. In this work, we use these maxims partially to measure the appropriateness or relevancy of an- swer(s) of a given question. In this approach, we do not try to understand what the speaker (inten- tional) means. Instead, we try to understand if the speaker contribution contains (extensional) el- ements that comply with Grice maxims. In the following, we explain how we interpret the quantity, relation and manner maxims in our approach. We do not use the quality maxim and it is beyond the scope of our research. 2.1 Quantity We interpret this maxim as how much an answer is informative by examining whether an answer con- tains the following informative elements? 1. Named …

Content truncated to 500 words