FA3L at SemEval-2017 Task 3: A ThRee Embeddings Recurrent Neural Network for Question Answering

Proceedings of the 11th International Workshop on Semantic Evaluations (SemEval-2017), pages 299–304, Vancouver, Canada, August 3 - 4, 2017. c⃝2017 Association for Computational Linguistics FA3L at SemEval-2017 Task 3: A ThRee Embeddings Recurrent Neural Network for Question Answering Giuseppe Attardi, Antonio Carta, Federico Errica, Andrea Madotto, Ludovica Pannitto Department of Computer Science, University of Pisa Largo B. Pontecorvo, 3 attardi@di.unipi.it, a.carta@outlook.it, f.errica@protonmail.com andreamad8@gmail.com, ellepannitto@gmail.com Abstract In this paper we present ThReeNN, a model for Community Question Answer- ing, Task 3, of SemEval-2017. The pro- posed model exploits both syntactic and semantic information to build a single and meaningful embedding space. Using a de- pendency parser in combination with word embeddings, the model creates sequences of inputs for a Recurrent Neural Network, which are then used for the ranking pur- poses of the Task. The score obtained on the official test data shows promising re- sults. 1 Introduction Community Question Answering (cQA) systems have proven to be useful for a long time and they still are an invaluable source of information. How- ever, due to their rapid growth and to the large amount of data provided it is not easy to find a rel- evant answer or a good related question amongst all the others. For these reasons we present a model which tries to tackle these problems. The subtasks we have worked on can be described as follows: A) Question-Comment Similarity - Given a question q and 10 comments c1, . . . , c10, rank such comments from the most relevant to the least one with respect to q, and assign to each one a label which can be "Good" or "Bad". B) Question-Question Similarity - Given a ques- tion q and a set of 10 related questions q1, . . . , q10, rank the 10 questions from "Relevant" to "Irrele- vant", according to q. A more detailed description of the task can be found in (Nakov et al., 2017). Our work has been inspired by studies regarding embedding spaces. Indeed, in (Hsu et al., 2016) GloVe embeddings (Pennington et al., 2014) are used to solve the same subtasks as ours, achieving good results using just word embeddings which encode semantic information into a vector. More- over, the model proposed in (Yu et al., 2013), where autoencoders are used to build an embed- ding space, has been exploited to propose an ap- proach that mixes semantic and syntactic informa- tion through the use of word embeddings and de- pendency parsing. These are then put together and become an input for the neural network. In this way we try to enhance the capability of the learn- ing system. In principle, our approach aims at enriching se- mantic information with syntactic relations hold- ing between elements of the couples (question- comment or question-question). This should serve well for both subtasks A and B, since the model will learn relations between a question and a com- ment or between a question and another one. How- ever, further research would be useful …

Content truncated to 500 words