Jamo Pair Encoding: Subcharacter Representation-based Extreme Korean Vocabulary Compression for Efficient Subword Tokenization

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 340–345 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 340 Creating Corpora for Research in Feedback Comment Generation Ryo Nagata, Kentaro Inui, Shin’ichiro Ishikawa Konan University/JST, Presto, Tohoku University/RIKEN AIP, Kobe University 8-9-1 Okamoto, Higashinada-ku, Kobe, Hyogo 658-8501, Japan 6-6-05 Aoba, Aramaki, Aoba-ku, Sendai, 980-8579, Japan 1-2-1 Tsurukabuto, Nada-ku, Kobe, Hyogo 657-8501, Japan nagata-lrec2019 @ ml.hyogo-u.ac.jp., inui@ecei.tohoku.ac.jp, aiskwshin@gmail.com Abstract In this paper, we report on datasets that we created for research in feedback comment generation — a task of automatically generating feedback comments such as a hint or an explanatory note for writing learning. There has been almost no such corpus open to the public and accordingly there has been a very limited amount of work on this task. In this paper, we first discuss the principle and guidelines for feedback comment annotation. Then, we describe two corpora that we have manually annotated with feedback comments (approximately 50,000 general comments and 6,700 on preposition use). A part of the annotation results is now available on the web, which will facilitate research in feedback comment generation. Keywords: Feedback comment generation, Writing learning, Grammatical error correction 1. Introduction This paper reports on datasets that we have created for re- search in feedback comment generation. Feedback com- ment generation is a task of generating feedback comments for a given text (hereafter, essay) automatically. A feed- back comment is a hint or an explanatory note for the writer (typically, a language learner) that helps them improve their writing skill. As shown in Fig. 1 and Fig. 2, it is typically a comment on erroneous, unnatural, or problematic words in a given text so that the writer can understand why the present form does not match with the underlying rule. The purpose of this corpus creation is to facilitate research in feedback comment generation. It will augment grammat- ical error correction as language learning assistance if one can automatically generate natural and effective feedback comments with grammatical error correction results. Unfortunately, however, there has been a very limited amount of work on feedback comment generation as Sect. 2. will describe. This is largely due to the fact that there had been no publicly available corpus annotated with feedback comments. Corpora annotated with feedback comments will likely facilitate the research in this domain just as the common datasets in grammatical error correction such as the CoNLL shared-task datasets (Ng et al., 2013; Ng et al., 2014) did. Recently, Nagata (2019) has released a dataset consisting of learner corpora manually annotated with feedback comments on preposition use; their target corpora are the two existing corpora — the written essays in ICNALE (Ishikawa, 2013) and KJ (Nagata et al., 2011). In this work, we extended the dataset by manually anno- tating the two learner corpora with feedback comments in general and those on preposition use (hereafter, general and preposition feedback comments, respectively). We also ex- tended guidelines for this annotation. …

Content truncated to 500 words