Received March 11, 2018, accepted April 20, 2018, date of publication May 1, 2018, date of current version May 24, 2018. Digital Object Identifier 10.1109/ACCESS.2018.2832137 Application of Sentiment Analysis to Language Learning MEI-HUA CHEN 1 , WEI-FAN CHEN 2 , AND LUN-WEI KU 3 1 Department of Foreign Languages and Literature, Tunghai University, Taichung 40704, Taiwan 2 Bauhaus-Universität Weimar, 99423 Weimar, Germany 3 Institute of Information Science, Academia Sinica, Taipei 11529, Taiwan Corresponding author: Mei-Hua Chen (mhchen@thu.edu.tw) This work was supported by the Ministry of Science and Technology, Taiwan, under Grant MOST 105-2410-H-029-041. ABSTRACT Emotion vocabulary has been studied in various disciplines, such as psychology, linguistics, and computational linguistics. Recently, it plays a requisite role in sentiment analysis or opinion mining. However, emotion vocabulary has not received considerable attention in second or foreign language learning. The insufficient pedagogical materials and inefficient tool support seem to provide little help for learners to master emotion words. The current study considers the application of sentiment analysis to language learning. To achieve this goal, we developed RESOLVE, a context-aware emotion synonym suggestion system, for educational purposes. Utilizing machine-learning techniques, the system is capable of suggesting synonymous emotion words appropriate to learners’ contexts. Importantly, the usage information of each emotion word, including scenario descriptions, definitions, and example sentences, is provided in order to help develop language learners’ vocabulary knowledge as well as help facilitate their word use. A pedagogical evaluation of the system’s effectiveness was conducted using a writing task and a survey questionnaire. The results indicate that the participants achieved substantial progress on emotion word use with the help of the proposed system. In particular, less proficient participants demonstrated greater improvements. Meanwhile, participants showed positive attitudes toward the tool support, as it helps them to have a better command of emotion words in their writings. INDEX TERMS Computer aided instruction, context awareness, educational technology, natural language processing. I. INTRODUCTION Emotion vocabulary has been studied in various disciplines, such as psychology [1], [2], linguistics, and computational linguistics [3]–[5]. In particular, emotion vocabulary has been an essential component in sentiment analysis or opinion min- ing, tasks of great interest within the field of Artificial Intel- ligence combining Natural Language Processing, Machine Learning and Psychology [6]–[8] over the past two decades. However, emotion vocabulary has received little atten- tion in second or foreign language learning [5], [9]–[12]. Emotion words are either absent in pedagogical materi- als [9], [13] or are not emphasized in classrooms [11], [14]. Little exposure leads to language learners’ limited vocab- ulary size [15]. As a result, learners opt for general terms or superordinates (e.g., ‘‘happy’’) rather than specific terms or hyponyms (e.g., ‘‘thrilled’’) to describe their feel- ings [15], [16]. On the other hand, the existing reference tools, such as dictionaries or thesauri, appear not directly or effectively to help learners discriminate the nuances of syn- onymous emotion words. The suggested emotion words may be accompanied by definitions but seldom carry explanations of how they could be used [16], [17]. As a result, even though learners learn the definitions of synonymous emotion words, they may still be unable to determine the appropriate word to ‘‘fit the concept being expressed’’ [18]. To address this issue, this study applied the sentiment anal- ysis techniques to assisting language learning. We developed RESOLVE (Ranking Emotional SynOnyms for language Learners’ Vocabulary Expansion) [19], [20], a context-aware emotion synonym suggestion system. Taking advantage of machine learning capabilities of learning from and making predictions on data, the RESOLVE system is able to suggest appropriate emotion synonyms based on learners’ contex- tual information. Importantly, each suggested emotion syn- onym comes along with usage information, including word definitions, scenario descriptions, and example sentences. VOLUME 6, 2018 2169-3536  2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. 24433