Yo Ehara

I have been studying learning support systems/natural language processing/computational linguistics with a focus on language learner support systems. Primary research interests and accomplishments as of December 2022 are summarized as follows:

1. The relationship between contextualized word embedding, such as BERT [Devlin et al., 2019], and word difficulty

• A study of extracting word difficulty from contextualized word embeddings without word-difficulty supervision by using outlier detection [Ehara, AIED 2022 (full paper)].
• A study of extracting learner ability values from contextualized word embeddings from vocabulary test result dataset of second language learners. [Ehara, EDM short2022].

2. Research on Automatic Readability Assessment (ARA) for Second Language Learners

• Unsupervised (in the sense that no language teachers' costly manual readability labels are used as supervision) Automatic Readability Assessment by using learners' vocabulary test result data [Ehara, ICTAI 2021].

Tenured senior lecturer on Natural Language Processing, especially educational NLP and second language vocabulary acquisition. Built publicly-available datasets on second language vocabulary knowledge: http://yoehara.com/vocabulary-prediction/ .