ECNU at SemEval-2018 Task 11: Using Deep Learning Method to Address Machine Comprehension Task

Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 1048–1052 New Orleans, Louisiana, June 5–6, 2018. ©2018 Association for Computational Linguistics ECNU at SemEval-2018 Task 11: Using Deep Learning Method to Address Machine Comprehension Task Yixuan Sheng1, 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 51164500026@stu.ecnu.edu.cn, {mlan, ybwu}@cs.ecnu.edu.cn Abstract This paper describes the system we submitted to the Task 11 in SemEval 2018, i.e., Machine Comprehension using Commonsense Knowl- edge. Given a passage and some questions that each have two candidate answers, this task re- quires the participate system to select out one answer meet the meaning of original text or commonsense knowledge from the candidate answers. For this task, we use a deep learning method to obtain final predict answer by calcu- lating relevance of choices representations and question-aware document representation. 1 Introduction In recent years, the presentation of challenge and large-scale reading comprehension corpora has driven the development of technology for machine reading comprehension, and most of these ma- chine comprehension datasets do not need com- monsense knowledge to answer questions. The purpose of Machine Comprehension using Com- monsense Knowledge task in Semeval 2018 is to provide a platform for finding a way for the machine to better understand the text and enable the machine answer questions based on the text, and encourage participants to make use any ex- ternal resources (e.g., DeScript, narrative chains, Wikipedia, etc) to improve the system perfor- mance (Ostermann et al., 2018b). The task 11 is a multiple-choice machine comprehension, which requires a system read a narrative text about every- day activities (Ostermann et al., 2018a) and then answer multiple-choice questions based on this text. Some questions need to be answered accord- ing to the original text, and others can be answered by commonsense knowledge. Each question is as- sociated with a set of two answers. Table 1 gives an example of the dataset. To address this machine comprehension task, we utilized rule-based methods and a deep learn- document: Early this morning I woke up to the sound of my newspaper landing on my driveway. I sat up and wrapped my pink robe around me. I slipped my feet into my slippers and looked at the clock. It was only 7:00 but it was time for me to get my newspaper and drink some coffee. I looked out the window and noticed it was raining quite a bit. I saw the newspaper at the end of my driveway, as far away as it could be. I grabbed my umbrella out of my coat closet and opened my front door enough to stick the umbrella through and open it outside. I stepped out the door and quickly covered my head with the umbrella. Then I ran to the end of my driveway, scooped up the newspaper in its plastic wrapping, and ran back to my front door. I closed my umbrella, took off my slippers, and dried …

Content truncated to 500 words