Compiling Learner Corpus Data of Linguistic Output and Language Processing in Speaking, Listening, Writing, and Reading

Proceedings of the 5th International Joint Conference on Natural Language Processing, pages 1418–1422, Chiang Mai, Thailand, November 8 – 13, 2011. c⃝2011 AFNLP Compiling Learner Corpus Data of Linguistic Output and Language Processing in Speaking, Listening, Writing, and Reading Katsunori Kotani Kansai Gaidai University / Osaka, Japan kkotani@kansaigaidai.ac.jp Takehiko Yoshimi Ryukoku University / Shiga, Japan yoshimi@rins.ryukoku.ac.jp Hiroaki Nanjo Ryukoku University / Shiga, Japan nanjo@rins.ryukoku.ac.jp Hitoshi Isahara Toyohashi University of Technology / Aichi, Japan isahara@tut.jp Abstract A learner’s language data of speaking, writing, listening, and reading have been compiled for a learner corpus in this study. The language data consist of linguistic output and language processing. Linguistic output refers to data of pronunciation, sentences, listening compre- hension rate, and reading comprehension rate. Language processing refers to processing time and learners’ self-judgment of their difficulty of processing in speaking, listening, and read- ing and the fluency of their writing. This learner corpus will contribute to making the language learning process more clearly visible. 1 Introduction Learner corpora have contributed to second lan- guage acquisition (SLA) research. For instance, SLA research using learner corpora examines learners’ proficiency on the basis of what voca- bularies/grammars learners actually use (Tono 2009, among others). Thus, most learner corpora are compiled of linguistic outputs that learners produce in speaking and/or writing. In order to enhance SLA research, a learner corpus should be compiled of a learner’s lan- guage data of the four modalities (speaking; lis- tening; writing; reading). Language data of each modality are further classified into two types: linguistic output and language processing data. Language output data consist of linguistic objects that learners produce. Language processing data indicate how they produce linguistic outputs. Thus, we have eight types of language data that are useful for the SLA research on the develop- ment of learners’ proficiency (Hinkel 2010, Sega- lowitz 2003). Among these eight types of lan- guage data, the previous studies (Granger et al. 2009, Izumi et al. 2004, Meurers et al. 2010, Wen et al. 2008) compiled the language output data of speaking, writing, and reading for constructing a learner corpus. See Section 2 for further detail. On the other hand, the other previous studies (Zechner & Bejar 2006, Arthur 1979, Hirai 1999, Kotani et al. 2010, Chang 2010) compiled the language processing data not for constructing a learner corpus but for examining learners’ per- formance. See Section 3 for further detail. Thus, there is a shortage of language output data of lis- tening, and furthermore we have to construct a learner corpus that integrates all these eight types of data. Hereafter, we refer to this corpus as I(ntegrated)-Learner Corpus. In order to construct I-Learner Corpus, we have compiled data of lin- guistic output and language processing of the four modalities when learners actually use the target language, which in this study is English. 2 Background Written data are compiled in the International Corpus of Learner English (ICLE) (Granger et al. 2009). The written data are taken from essays written by learners of English as …

Content truncated to 500 words