Grid Tagging Scheme for Aspect-oriented Fine-grained Opinion Extraction

Findings of the Association for Computational Linguistics: EMNLP 2020, pages 2576–2585 November 16 - 20, 2020. c⃝2020 Association for Computational Linguistics 2576 Grid Tagging Scheme for Aspect-oriented Fine-grained Opinion Extraction Zhen Wu1 Chengcan Ying1 Fei Zhao1 Zhifang Fan1 Xinyu Dai1 ∗ Rui Xia2 1National Key Laboratory for Novel Software Technology, Nanjing University 2School of Computer Science and Engineering, Nanjing University of Science and Technology {wuz, yingcc, zhaof, fanzf}@smail.nju.edu.cn daixinyu@nju.edu.cn, rxia@njust.edu.cn Abstract Aspect-oriented Fine-grained Opinion Extrac- tion (AFOE) aims at extracting aspect terms and opinion terms from review in the form of opinion pairs or additionally extracting senti- ment polarity of aspect term to form opinion triplet. Because of containing several opinion factors, the complete AFOE task is usually di- vided into multiple subtasks and achieved in the pipeline. However, pipeline approaches easily suffer from error propagation and incon- venience in real-world scenarios. To this end, we propose a novel tagging scheme, Grid Tag- ging Scheme (GTS), to address the AFOE task in an end-to-end fashion only with one uni- fied grid tagging task. Additionally, we de- sign an effective inference strategy on GTS to exploit mutual indication between differ- ent opinion factors for more accurate extrac- tions. To validate the feasibility and compat- ibility of GTS, we implement three different GTS models respectively based on CNN, BiL- STM, and BERT, and conduct experiments on the aspect-oriented opinion pair extraction and opinion triplet extraction datasets. Extensive experimental results indicate that GTS models outperform strong baselines significantly and achieve state-of-the-art performance. 1 Introduction Aspect-oriented Fine-grained Opinion Extraction (AFOE) aims to automatically extract opinion pairs (aspect term, opinion term) or opinion triplets (as- pect term, opinion term, sentiment) from review text, which is an important task for fine-grained sentiment analysis (Pang and Lee, 2007; Liu, 2012). In this task, aspect term and opinion term are two key opinion factors. Aspect term, also known as opinion target, is the word or phrase in a sentence representing feature or entity of products or ser- vices. Opinion term refers to the term in a sentence ∗† Corresponding author. Sentence: The hot dogs are top notch and great coffee Aspect Terms: hot dogs, coffee Opinion Terms: top notch, great Opinion Pairs: (hot dogs, top notch), (coffee, great) Opinion Triplets: (hot dogs, top notch, positive), (coffee, great, positive) Sentence: The hot dogs are top notch but average coffee Aspect Terms: Opinion Terms: Opinion Pairs: Opinion Triplets: hot dogs, coffee top notch, average (hot dogs, top notch), (coffee, average) (hot dogs, top notch, positive), (coffee, average, neutral) Figure 1: An example of aspect-oriented fine-grained opinion extraction. The spans highlighted in red are aspect terms. The terms in blue are opinion terms. used to express attitudes or opinions explicitly. For example, in the sentence of Figure 1, “hot dogs” and “coffee” are two aspect terms, “top notch” and “average” are two opinion terms. To obtain the above two opinion factors, many works devote to the co-extraction of aspect term and opinion term in a joint framework (Wang et al., 2016, …

Content truncated to 500 words