Leveraging a Large Learner Corpus for Automatic Suggestion of Collocations for Learners of Japanese as a Second Language

Oct 15, 2026

Learning Japanese as a second language often requires selecting word combinations that sound natural, yet learners lack the native-like experience to judge which collocations fit specific usage. This study builds a collocational aid that generates and ranks noun–verb correction candidates. Candidates are extracted from a large Japanese learner corpus, filtered using the Weighted Dice association measure, and evaluated on learner data. The system shows statistically stronger performance than existing collocation error correction approaches and includes a preliminary user study with JSL learners.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages23

Reading Time58 min

AccessFree to download

Introduction

Collocations and their role in second language learning

Limitations of existing tools for collocation errors

Rationale for an automatic collocation error correction system

calico journal (online) issn 2056–9017 Article Leveraging a Large Learner Corpus for Automatic Suggestion of Collocations for Learners of Japanese as a Second Language Lis Pereira, Erlyn Manguilimotan, and Yuji Matsumoto Abstract One of the challenges of learning Japanese as a Second Language (JSL) is finding the appropriate word for a particular usage. To address this challenge, we developed a collocational aid designed to suggest more appropriate collocations in Japanese. In particular, we address the problem of generating and ranking noun and verb candidates for correcting potential collocation errors in the learners’ text. Given a noun–verb construction as input, our system generates possible noun or verb correction candidates based on noun and verb corrections extracted from a large Japanese learner corpus. We use this corpus to investigate the learner’s tendency to commit collocation errors, and to produce a smaller and more realistic set of candidates. After combining nouns or verbs with the generated candidates to form noun–verb pairs, the system uses the Weighted Dice coefficient as the association measure to filter out inappropriate noun–verb pairs and rank the proper collocations. We report the detailed evaluation and results on learner data. In addition, we show that our system statistically outperforms existing approaches to collocation error correction. Finally, we report a preliminary user study with JSL learners. Keywords: Collocation; Automatic Error Correction; Language Learner; Japanese Affiliation Nara Institute of Science and Technology. email: [lis-k@is.naist.jp](lis-k@is.naist.jp) (corresponding author) calico journal vol 33.3 2016 311–333 ©2016, equinox publishing doi : 10.1558/cj.v33i3.26444 312 Leveraging a Large Learner Corpus 1. Introduction One of the challenges of learning a second language is finding the appropriate word for a particular usage. Learners of a second language do not yet have the extensive experience of native speakers to know which words are often combined to make natural expressions. The accurate use of words that commonly occur together, or simply collocations, is crucial for clear and effective communication similar to that of a native speaker. Lewis (2000) argues that“increasing the learners’ collocational competence is the way to improve their language as a whole” (p. 14) . In a separate study, Hill (2000, p. 62) explains that a student who uses collocations competently will be far more competent in communication than a student who does not. The literature defines collocations in several ways. Lea and Runcie define collocations as combinations of words in a language to produce natural-sounding speech and writing (2002, p. vii) . Smadja describes collocations as recurrent combinations of words that co-occur with higher possibility than random chance, and correspond to some arbitrary word usages (1993, p. 143) . Regarding the semantic compositionality, collocations can be distinguished from idioms, which have meanings that are more opaque (Seretan, 2011, p. 23) . Despite being less fixed compared to idioms, collocations would be regarded as less appropriate when one of the components is replaced by another word (Chang, Chang, Chen, & Liou, 2008; Shei, & Pain, 2000; Leacock, Chodorow, Gamon, & Tetreault, 2010, p. 65). In summary, for the purposes ofour present study, collocations are word combinations that: 1. are arbitrary and recurrent; [2. co-occur more often than expected by chance](2. co-occur more often than expected by chance); and 3. would be regarded as less appropriate when one of the components is replaced by another word. Studies confirm that the correct use of collocations is challenging, even for advanced second language learners (Liu, 2002; Nesselhauf, 2003; Wible, Kuo, Tsao, Liu, & Lin, 2003). Unfortunately, the number of tools designed to target language learner collocation errors is limited. Most spell checkers and grammar che

FAQs

What problem does the paper address for learners of Japanese as a second language?

It addresses the difficulty of choosing appropriate word combinations (collocations) for specific usage in Japanese, especially when correcting potential noun–verb collocation errors in learner writing.

How does the proposed system generate correction candidates?

Given a noun–verb construction, the system generates possible noun or verb correction candidates using noun and verb corrections extracted from a large Japanese learner corpus, then forms noun–verb pairs.

What method does the system use to rank collocations and filter inappropriate pairs?

It uses the Weighted Dice coefficient as an association measure to filter out inappropriate noun–verb pairs and rank the proper collocations, followed by detailed evaluation on learner data.

Oct 15, 2026

Uploaded by Well Done