Basic study for Japanese Composition Support System for Scientific Students employing Natural Language Processing Techniques.

我们为第二语言学习者从三个不同的角度研究了日语的构成。1)利用自然语言处理技术：我们开发了一个日语名词和形容词共现词的提取工具。2)学习者偏误分析：使用开发的分析工具，我们创建了一个由中国学生撰写的日语作文语料库，分析了作文中名词和形容词的共现情况。为了找出中国学习者的错误模式，我们与共现抽取工具提取的标准日语用法的例子进行了比较。基于这些对比，我们发现中国学习者在使用NA形容词时容易犯错误，尤其是在汉语中也存在但在日语中用法不同的词。3)建立连贯的科技作文语篇为了实现理工科学生写作的综合课程，一个中心任务是呈现科技论文的整体结构。为此，我们分析了科技文献语篇文体的特点，并从工程和农业期刊文章的前言和结束语中提取了一些典型的句型。这一分析清楚地强调了向学习者展示写作图式的重要性，并补充了尽可能多的适当例子。

We have studied Japanese composition from three different perspectives for second language learners.1) Utilizing natural language processing techniques :We have developed an extraction tool that presents some co-occurrence words for Japanese nouns and adjectives. We extracted co-occurrence data for adjectives and nouns from newspaper articles covering a total of 10 years using morphological and syntactic analyzers.2) Learner error analysis : Employing a developed analysis toolWe created, a corpus of Japanese compositions written by Chinese student, which we analyzed for co-occurrence of nouns and adjectives within the compositions. In order to identify the error patterns for Chinese learners, we made comparisons with examples of standard Japanese usage extracted by the co-occurrence extraction tool. Based on these comparisons, we found that Chinese learner tend to make errors with na-Adjectives, particular for words that also exist in Chinese but where the usage is different in Japanese.3) Creating coherent discourse for scientific compositionsIn order to realize comprehensive courses of written composition for technical students, a central task is to present the overall structures of scientific papers. To that aim, we have analyzed the specific features of discourse styles in scientific documents.We extracted a number of typical sentence patterns from the prefaces and result sections of engineering and agricultural journal articles. This analysis clearly highlighted the importance of presenting writing schema to learners with a rich supplement of as many appropriate examples as possible.