Appropriate lexical use is one daunting task for many language learners even though they have a wide range of vocabulary. It is particularly true when they express their emotions. With inadequate command of emotion words, learners tend to use common emotion words (e.g., “angry” or “happy”) to describe their feelings. Sometimes, they attempt to use alternative words; they would consult thesauri for synonyms lookup. However, the synonyms thesauri suggest are typically listed in alphabetical order and short of contextual information, which seem unable to help learners tell the nuanced emotion words. As a result, the synonyms learners choose are very likely to fail to fit the scenarios they would like to describe. Bearing this in mind, we utilized machine learning technique to develop an emotion wording assistance system, RESOLVE to help learners with their emotional expressions. More specifically, the system suggests appropriate synonymous emotion words which are ranked based on learners’ contexts. In addition, the corresponding usage information involving the description of the usage scenarios, definitions and example sentences is also provided. Such information aims to facilitate learners’ clear description of their emotions.
To evaluate the effectiveness of the RESOLVE system, we carried out an experiment with 36 EFL college students in an Asian country. We compared their writing tasks in the pre- and post-tests to examine the appropriateness of emotion wording and the difficulty of emotion words students achieved. Two native English speaker judges were involved to evaluate students’ performance. The results showed that with the help of RESOLVE, all students achieved substantial improvements in appropriate emotion wording (i.e., the average scores increased from 56.6 to 88.3 out of 100.0). Importantly, the less proficient gained more benefit from RESOLVE than high proficient ones (the respective error reduction rates are 76.2% and 49.0%). On the other hand, the difficulty level of emotion words students used also increased from 51.5% to 67.7%, inferring that students had a better command of emotion words. In addition, students’ attitudes toward RESOLVE were predominantly positive.
