Multilingual coprus of program and its document-from the viewpoint of "Software = program + document"-

我们的目标是使CASE工具既能处理用人工语言编写的源程序，又能处理用自然语言描述的文档。我们的研究结果如下：1。我们开发了源程序中的注释模型，作为C语言细粒度模型的扩展。我们提出了一种利用在源程序和文档中同时出现的标识符来获取源程序和文档之间对应关系的方法。我们的方法通过考虑上下文来区分“定义”和“参考”。通过分析从源程序中提取的评论句子的结构，提出了一种揭示评论描述特征的方法。提出了一种对日语技术文档中频繁出现的假名复合词进行分类的方法。我们提出了一个CASE工具平台支持环境，它可以被一组编程语言重用。我们开发了一个帮助标识符命名的系统，延迟了项目的定制。我们系统的目标是Java的方法名。系统从现有的方法名及其文档注释中提取命名自定义。我们提出了一种帮助用英语写评论的方法。该方法使用现有注释作为语料库。我们提出了一种自动识别关键词序列结构模式作为查询的语料库搜索方法，目的是利用我们的语料库对语料库进行检验和提炼。

We aim the CASE tool that handles not only source program written in artificial language but also document described in natural language. The results of our research are as follows:1. We developed the model of comments in source program, as an extension of our fine grained model of C language.2. We proposed a method to take correspondence between source program and document, by using identifiers which appear in both. Our method distinguishes "definition" and "reference" by considering context.3. We proposed a method to disclose the characteristics of comment description, by analyzing structure of comment sentences extracted from source program.4. We proposed a method to divide compound words of KANA, which appear frequently in technical document in Japanese.5. We proposed a CASE tool platform support environment which is reusable for a group of programming languages.6. We developed a system that assists identifier naming, deferring the custom of the project. The target of our system is method name of Java. The system extracts the naming custom from existing method names and its documentation comment.7. We proposed a method to assist writing comments in English. The method uses existing comments as the corpus.8. We proposed a corpus search method that automatically identifies structural patterns of sequences of keywords as query, aiming to examine and refine the corpus, by making application of our corpus.