This paper is part of a larger research project aimed at building a composition support system for L2 Japanese learners writing academic reports, graduation theses, and academic papers. As an extension to previous research on conjunctive expressions, this research examines the cooccurrence relationship between conjunctive expressions and sentence-final modality forms using a variety of written corpora including research papers, textbooks, a subset of the Balanced Corpus of Contemporary Written Japanese, and several academic learner corpora. First, we quantitatively analyze the cooccurrence distributions of conjunctive expression and sentence-final modality form pairs, and then extend the analysis to also include sentence-final modality forms from the preceding sentence. By focusing on the most salient and typical collocations using a combination of corpus-normalized frequencies, pointwise mutual information, and entropy, and observing them from the perspective of discourse analysis, we uncover the characteristics of discourse structure in the genre of academic writing and compare them with those of L2 learner academic writing. Through the process of identifying problems in the discourse structure of learner corpora compared to reference academic corpora, we aim to help learners write conjunctive expressions and sentence-final modality forms in the appropriate academic style.

View full abstract

The present research reports on an analysis of essays on the themes of ‘past’ and ‘future’. We assign the UniDic morphological information to the digitized essays and analyze them in terms of part-of-speech realization frequencies. Then we match the identified vocabulary with the Fundamental Vocabulary of Educational Purpose and analyze them in terms of the occurrence trend of vocabulary allocation. We also analyzed the data based on the Generalized Linear Mixed Model and compared them by the essay themes and the author attributes such as age groups and gender types. The results have revealed how each attribute characterizes the essay style of each group. The differences are found in the frequency of the “tai (nouns)”, “yoo (verbs)”, and “soo (modifiers)” as well as in the occurrence trend of vocabulary allocation.

View full abstract

This paper presents several application examples of reciprocal averaging on cross-tabulation analysis. First, examples of applying reciprocal averaging to extended cross-tab are discussed. Reciprocal averaging can be applied not only in a case where a hypothesis has been formulated and data are collected by questionnaire to confirm the hypothesis but also as a means of exploratory data analysis in a case without a specific hypothesis. It can also be applied to a type of questionnaire asking for scale points. Even for those with no scales in the answer section, the extended cross-tab obtained with its application would look easier to understand; several examples are given in this section to demonstrate. Next, reciprocal averaging application examples to past research are discussed to confirm how the application would make the results more understandable. Examples include a case of oversight found in the paper's description, a case of dealing with something not included in the paper's analysis subjects, a case of improving result presentation by modifying graphic charts initially organized by the author, and other relevant cases. Lastly, examples to demonstrate that applying reciprocal averaging would be meaningless in the specific type of cross-tabs are given to make the point that reciprocal averaging is not a universal tool enabling analysis of any kind of cross-tabs.

View full abstract