Solving Aspect Category Sentiment Analysis as a Text Generation Task

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 4406–4416 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 4406 Solving Aspect Category Sentiment Analysis as a Text Generation Task Jian Liu1, Zhiyang Teng2,3, Leyang Cui2,3, Hanmeng Liu2,3, Yue Zhang2,3 1School of Computer Science, Fudan University 2School of Engineering, Westlake University 3Institute of Advanced Technology, Westlake Institute for Advanced Study jianliu17@fudan.edu.cn, {tengzhiyang, cuileyang, liuhanmeng, zhangyue}@westlake.edu.cn Abstract Aspect category sentiment analysis has at- tracted increasing research attention. The dominant methods make use of pre-trained language models by learning effective aspect category-specific representations, and adding specific output layers to its pre-trained repre- sentation. We consider a more direct way of making use of pre-trained language models, by casting the ACSA tasks into natural language generation tasks, using natural language sen- tences to represent the output. Our method al- lows more direct use of pre-trained knowledge in seq2seq language models by directly follow- ing the task setting during pre-training. Exper- iments on several benchmarks show that our method gives the best reported results, having large advantages in few-shot and zero-shot set- tings. 1 Introduction Aspect-based sentiment analysis (ABSA) is a fine- grained sentiment analysis task that includes a num- ber of subtasks, two of which are aspect category sentiment analysis (ACSA) and aspect category de- tection (ACD). Figure 1 shows an example, where the input is “The restaurant was expensive, but the menu was great”. ACD detects the aspect cate- gories, such as price and food, and ACSA predicts the sentiment polarities toward each aspect cate- gory. In this work, we focus on these two tasks as well as the joint task that combines both. Previous studies have investigated various meth- ods that treat ACSA and ACD as classification tasks, learning aspect-specific sentence representa- tions (Wang et al., 2016; Ruder et al., 2016). Re- cently, pre-trained language models (PLM) have shown their effectiveness to this end (Jiang et al., 2019). The main idea is to make use of pre-trained models such as BERT (Devlin et al., 2019a) for rep- resenting an aspect-specific form of the input (e.g., by concatenating the aspect category to the end of The restaurant was expensive, but the menu was great <price, food> < price: negative > < food: positive > ACSA ACD Figure 1: Example of aspect category detection (ACD) and aspect category sentiment analysis (ACSA). the input sentence (Figure 3(a))), which provides useful semantic features for ACSA and ACD classi- fiers. Such methods have given highly competitive results (Sun et al., 2019; Li et al., 2020b). The above classification models benefit from contextualized representations, which contain knowledge learned by pre-training over large data (Lin et al., 2019). However, their use of pre-trained knowledge can be viewed as indirect due to at least two reasons. First, the classification task is per- formed by using a neural network on top of pre- trained representation, with separate network pa- rameters. Second, the integration of aspect cate- gory makes the aspect-specific input representation not exactly a natural language …

Content truncated to 500 words