Semi-Supervised Learning Based on Auto-generated Lexicon Using XAI in Sentiment Analysis

Proceedings of Recent Advances in Natural Language Processing, pages 593–600 Sep 1–3, 2021. https://doi.org/10.26615/978-954-452-072-4_067 593 Abstract In this study, we proposed a novel Lexicon-based pseudo-labeling method utilizing explainable AI(XAI) approach. Existing approach have a fundamental limitation in their robustness because poor classifier leads to inaccurate soft-labeling, and it lead to poor classifier repetitively. Meanwhile, we generate the lexicon consists of sentiment word based on the explainability score. Then we calculate the confidence of unlabeled data with lexicon and add them into labeled dataset for the robust pseudo-labeling approach. Our proposed method has three contributions. First, the proposed methodology automatically generates a lexicon based on XAI and performs independent pseudo- labeling, thereby guaranteeing higher performance and robustness compared to the existing one. Second, since lexicon- based pseudo-labeling is performed without re-learning in most of models, time efficiency is considerably increased, and third, the generated high-quality lexicon can be available for sentiment analysis of data from similar domains. The effectiveness and efficiency of our proposed method were verified through quantitative comparison with the existing pseudo-labeling method and qualitative review of the generated lexicon. 1 Introduction Sentiment analysis is employed to identify the sentiment orientation and measure the emotional strength (Khan et al., 2016; Khan & Lee, 2019; Silva et al., 2016). To better understand information generated by online user and to take advantage of it, sentiment analysis is becoming major topic in text mining field in last two decades (Duan et al., 2020; Nagarajan & Gandhi, 2019; Valdivia et al., 2017). Previous studies of sentiment analysis can be roughly categorized into two different groups: 1) lexicon-based approaches and 2) machine learning-based approaches (Khan et al., 2019; Khoo & Johnkhan, 2018). The lexicon-based approaches efficiently calculate the sentiment score of sentence of document since it does not need to train the classification model in advance. However, The lexicon-based approaches depends on the availability of a sentiment lexicon which is collection of manually pre-created sentiment words lexicon and its sentiment polarity (Huang et al., 2020; Taj et al., 2019; Alqaryouti et al., 2019). Meanwhile, the machine-learning based approaches require a training set consists of labeled data (e.g. positive, negative or neutral). To address the challenge caused by limited labeled data, which is the usual case in practice, semi-supervised learning have attracted more attention recently (Han et al., 2020; Zhang et al., 2020; Lee et al., 2019). The semi-supervised learning can be divided into several categories such as consistency regularization approaches, entropy minimization approaches and augmentation based approaches and pseudo-labeling approaches. Among those approaches, the pseudo-labeling approaches are one of the most intuitive and widely used semi-supervised learning in sentiment analysis (Xu & Tan, 2019; Wu et al., 2019; Chen et al., 2020). The pseudo-labeling approaches tried to train the sentiment classification model with small number of labeled data and add unlabeled data with high-confidence of sentiment score, calculated by Semi-Supervised Learning based on Auto-generated Lexicon using XAI in Sentiment Analysis Hohyun Hwang, Younghoon Lee Seoul National University of Science and Technology, Korea, South (KR) …

Content truncated to 500 words