Improving Federated Learning for Aspect-based Sentiment Analysis via Topic Memories

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 3942–3954 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 3942 Improving Federated Learning for Aspect-based Sentiment Analysis via Topic Memories Han Qin♠∗, Guimin Chen♦∗, Yuanhe Tian♥∗, Yan Song♠† ♠The Chinese University of Hong Kong (Shenzhen) ♦QTrade ♥University of Washington ♠hanqin@link.cuhk.edu.cn ♦chenguimin@foxmail.com ♥yhtian@uw.edu ♠songyan@cuhk.edu.cn Abstract Aspect-based sentiment analysis (ABSA) pre- dicts the sentiment polarity towards a par- ticular aspect term in a sentence, which is an important task in real-world applications. To perform ABSA, the trained model is re- quired to have a good understanding of the contextual information, especially the partic- ular patterns that suggest the sentiment polar- ity. However, these patterns typically vary in different sentences, especially when the sen- tences come from different sources (domains), which makes ABSA still very challenging. Al- though combining labeled data across different sources (domains) is a promising solution to address the challenge, in practical applications, these labeled data are usually stored at differ- ent locations and might be inaccessible to each other due to privacy or legal concerns (e.g., the data are owned by different companies). To address this issue and make the best use of all labeled data, we propose a novel ABSA model with federated learning (FL) adopted to overcome the data isolation limitations and incorporate topic memory (TM) proposed to take the cases of data from diverse sources (do- mains) into consideration. Particularly, TM aims to identify different isolated data sources due to data inaccessibility by providing use- ful categorical information for localized pre- dictions. Experimental results on a simulated environment for FL with three nodes demon- strate the effectiveness of our approach, where TM-FL outperforms different baselines includ- ing some well-designed FL frameworks.1 1 Introduction Aspect-based sentiment analysis (ABSA) is one of the most popular natural language processing *Equal contribution. †Corresponding author. 1The code involved in this paper are released at https: //github.com/cuhksz-nlp/ASA-TM. (NLP) tasks aiming to predict the sentiment po- larity (i.e., “positive”, “negative”, and “neutral”) for an aspect term in sentences. Currently, meth- ods based on deep learning have been widely uti- lized for ABSA and demonstrated excellent poten- tials (Chen et al., 2017; Zadeh et al., 2017; Zhang et al., 2018; Xue and Li, 2018; Zhao et al., 2018; Chaturvedi et al., 2018; Xu et al., 2019b). However, these methods still reach a bottleneck if there is no enough labeled training data. One feasible solution for it is to leverage extra labeled data from other sources or domains. However, in real applications, these data are always stored in different locations (nodes) and are inaccessible to each other owing to privacy or legal concerns. To address the data isolation issue, feder- ated learning (FL) (Shokri and Shmatikov, 2015; Koneˇcn`y et al., 2016a,b) is proposed and has shown its great promises for many machine learning tasks, such as user-computer interaction (Aono et al., 2017), medical image analysis (Sheller et al., 2018), and financial data analysis (Yang et al., 2019a; He et al., 2020). In …

Content truncated to 500 words