Argument Mining for Review Helpfulness Prediction

Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 926 - 941 December 7-11, 2022 ©2022 Association for Computational Linguistics MAVEN-ERE: A Unified Large-scale Dataset for Event Coreference, Temporal, Causal, and Subevent Relation Extraction Xiaozhi Wang1∗, Yulin Chen2∗, Ning Ding1, Hao Peng1, Zimu Wang5, Yankai Lin6,7† , Xu Han1, Lei Hou1, Juanzi Li1,3‡ , Zhiyuan Liu1, Peng Li4†, Jie Zhou8 1Department of Computer Science and Technology, BNRist; 2Shenzhen International Graduate School; 3THU-Siemens Ltd., China Joint Research Center for Industrial Intelligence and IoT; 4Institute for AI Industry Research (AIR), Tsinghua University, Beijing, China 5Xi’an Jiaotong-Liverpool University, Suzhou, China 6Gaoling School of Artificial Intelligence, Renmin University of China, Beijing, China 7Beijing Key Laboratory of Big Data Management and Analysis Methods, Beijing, China 8Pattern Recognition Center, WeChat AI, Tencent Inc, China {wangxz20,yl-chen21}@mails.tsinghua.edu.cn Abstract The diverse relationships among real-world events, including coreference, temporal, causal, and subevent relations, are fundamental to un- derstanding natural languages. However, two drawbacks of existing datasets limit event re- lation extraction (ERE) tasks: (1) Small scale. Due to the annotation complexity, the data scale of existing datasets is limited, which cannot well train and evaluate data-hungry models. (2) Absence of unified annotation. Different types of event relations naturally interact with each other, but existing datasets only cover limited relation types at once, which prevents mod- els from taking full advantage of relation in- teractions. To address these issues, we con- struct a unified large-scale human-annotated ERE dataset MAVEN-ERE with improved an- notation schemes. It contains 103, 193 event coreference chains, 1, 216, 217 temporal re- lations, 57, 992 causal relations, and 15, 841 subevent relations, which is larger than exist- ing datasets of all the ERE tasks by at least an order of magnitude. Experiments show that ERE on MAVEN-ERE is quite challeng- ing, and considering relation interactions with joint learning can improve performances. The dataset and source codes can be obtained from https://github.com/THU-KEG/MAVEN-ERE. 1 Introduction Communicating events is a central function of hu- man languages, and understanding the complex relationships between events is essential to under- standing events (Levelt, 1993; Miller and Johnson- Laird, 2013; Pinker, 2013). Thus event relation ∗ indicates equal contribution. † Partly done while Y.Lin and P.Li were at Tencent. ‡ Corresponding author: J.Li. Dataset #Doc. #Event Coref. #T-Link #C-Link #Subevent ACE 2005 599 4, 090 � � � � TAC KBP 1, 075 19, 257 � � � � TB-Dense 36 1, 712 � 10, 750 � � MATRES 275 11, 861 � 13, 573 � � Causal-TB 183 6, 811 � 5, 118 318 � EventStoryLine 258 4, 732 � 8, 111 4, 584 � HiEve 100 2, 734 � � � 3, 648 RED 95 2, 049 � 4, 209 1, 147 729 MAVEN-ERE 4, 480 103, 193 � 1, 216, 217 57, 992 15, 841 Table 1: Comparisons between MAVEN-ERE and most widely-used event relation datasets. T-Link and C- Link denote temporal and causal relations, respectively. #Event is the number of events (coreference chains) for the datasets …

Content truncated to 500 words