Implicit Sentiment Analysis with Event-centered Text Representation

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 6884–6893 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 6884 Implicit Sentiment Analysis with Event-Centered Text Representation Deyu Zhou1⇤, Jianan Wang1, Linhai Zhang1, Yulan He2 3 1School of Computer Science and Engineering, Key Laboratory of Computer Network and Information Integration, Ministry of Education, Southeast University, China 2Department of Computer Science, University of Warwick, UK 3Alan Turing Institute, UK {d.zhou,jnwang,lzhang472}@seu.edu.cn, yulan.he@warwick.ac.uk Abstract Implicit sentiment analysis, aiming at detect- ing the sentiment of a sentence without senti- ment words, has become an attractive research topic in recent years. In this paper, we focus on event-centric implicit sentiment analysis that utilizes the sentiment-aware event contained in a sentence to infer its sentiment polarity. Most existing methods in implicit sentiment analy- sis simply view noun phrases or entities in text as events or indirectly model events with so- phisticated models. Since events often trigger sentiments in sentences, we argue that this task would benefit from explicit modeling of events and event representation learning. To this end, we represent an event as the combination of its event type and the event triplet <subject, predi- cate, object>. Based on such event representa- tion, we further propose a novel model with hierarchical tensor-based composition mech- anism to detect sentiment in text. In addi- tion, we present a dataset1 for event-centric implicit sentiment analysis where each sen- tence is labeled with the event representation described above. Experimental results on our constructed dataset and an existing benchmark dataset show the effectiveness of the proposed approach. 1 Introduction Sentiment analysis aims at automatically detecting the sentiment of given text. Explicit sentiment anal- ysis methods detect the sentiment mainly based on the occurrence of sentiment-related words and have been extensively explored (Agarwal et al., 2011; Tang et al., 2014, 2015a). However, senti- ment could also be implicitly expressed. For exam- ple, the sentence ‘I won the first place in the speech contest’ does not contain any sentiment words, but the event of ‘winning the first place’ reflects the positive sentiment. Implicit sentiment analysis was ⇤Corresponding author. 1https://github.com/FloatingIsland2/ Implicit-Sentiment-Analysis thus proposed to detect sentiment in the absence of sentiment words (Liu, 2012). Compared with explicit sentiment analysis, implicit sentiment anal- ysis is more challenging as there is a lack of explicit cues for inferring the sentiment polarity. Research on implicit sentiment analysis can be broadly classified into two categories: metaphor- based and event-centric. Metaphor/rhetoric-based implicit sentiment analysis methods typically de- tect sentiment based on a metaphoric sentiment dic- tionary and some manually designed rules (Zhang and Liu, 2011), while event-centric approaches as- sume sentiment is triggered by events described in sentences. Examples include an early approach which regarded noun phrases or entities in sen- tences as events (Greene and Resnik, 2009), and more recent approaches which indirectly model events by capturing contextual information using graph convolutional networks (Zuo et al., 2020) or attention mechanism (Wei et al., 2020). We argue that in the former, the event representation is oversimplified …

Content truncated to 500 words