Syntactic and Semantic-driven Learning for Open Information Extraction

Findings of the Association for Computational Linguistics: EMNLP 2020, pages 54–64 November 16 - 20, 2020. c⃝2020 Association for Computational Linguistics 54 GRACE: Gradient Harmonized and Cascaded Labeling for Aspect-based Sentiment Analysis Huaishao Luo1∗, Lei Ji2,4,5, Tianrui Li1†, Nan Duan2, Daxin Jiang3 1School of Information Science and Technology, Southwest Jiaotong University, China huaishaoluo@gmail.com, trli@swjtu.edu.cn 2Microsoft Research Asia, Beijing, China 3STCA NLP Group, Microsoft, Beijing, China 4Institute of Computing Technology, Chinese Academy of Science, Beijing, China 5University of Chinese Academy of Sciences, Beijing, China {leiji,nanduan,djiang}@microsoft.com Abstract In this paper, we focus on the imbalance issue, which is rarely studied in aspect term extrac- tion and aspect sentiment classification when regarding them as sequence labeling tasks. Be- sides, previous works usually ignore the inter- action between aspect terms when labeling po- larities. We propose a GRadient hArmonized and CascadEd labeling model (GRACE) to solve these problems. Specifically, a cascaded labeling module is developed to enhance the interchange between aspect terms and improve the attention of sentiment tokens when label- ing sentiment polarities. The polarities se- quence is designed to depend on the gener- ated aspect terms labels. To alleviate the im- balance issue, we extend the gradient harmo- nized mechanism used in object detection to the aspect-based sentiment analysis by adjust- ing the weight of each label dynamically. The proposed GRACE adopts a post-pretraining BERT as its backbone. Experimental results demonstrate that the proposed model achieves consistency improvement on multiple bench- mark datasets and generates state-of-the-art re- sults. 1 Introduction Aspect terms extraction (ATE) and aspect senti- ment classification (ASC) are two fundamental, fine-grained subtasks in aspect-based sentiment analysis (ABSA). ATE is the task of extracting the aspect terms (or attributes) of an entity upon which opinions have been expressed, and ASC is the task of identifying the polarities expressed on these ex- tracted terms in the opinion text (Hu and Liu, 2004). Consider the example in Figure 1, which contains comments that people expressed about the aspect terms “operating system” and “keyboard”, and their polarities are all positive. ∗Work is done during an internship at MSR Asia. †Correspongding author. Input nice operating system and keyboard Joint O B I O B O POS POS O POS Collapsed O B-POS I-POS O B-POS Figure 1: Joint and Collapsed labeling approaches on aspect terms and their polarities. POS means positive. (a) Label statistics (b) Gradient statistics Figure 2: Label statistics and gradient distribution on the laptop dataset of SemEval-14. The y-axis in (b) uses a log scale. To better satisfy the practical applications, the aspect term-polarity co-extraction, which solves ATE and ASC simultaneously, receives much at- tention in recent years (Li et al., 2019b; Luo et al., 2019b; Hu et al., 2019; Wan et al., 2020). A big challenge of the aspect term-polarity co-extraction in a unified model is that ATE and ASC belong to different tasks: ATE is usually a sequence labeling task, and ASC is usually a classification task. Pre- vious works usually transform the ASC task into sequence labeling. Thus …

Content truncated to 500 words