Development of an Integrated Tool System for Technical Term Auto-Extraction and Knowledge Acquisition from Corpora

该项目的目标是提供能够以半自动方式从文本中获取术语知识的系统。为了实现这一目标，我们开发了以下三个系统：1.术语中央数据库：将EDR开发的文本/词汇数据库与东京大学开发的程序设计语言LiLFeS相结合，创建了一个术语数据库系统，方便灵活地处理语言实体通过该系统，我们可以对以下两个系统获得的知识进行系统维护。2.术语识别系统：NACSIS的研究小组引入了一种统计指标来识别文本中的技术术语，并构建了可以使用该指标识别术语的程序。东京大学的研究小组从不同的角度解决了同样的问题，并成功地提供了一种基于字符n-gram的术语识别方法。这些程序是集成的，因此它们可以作为1.3中描述的数据库系统的前端工作。获取术语本体知识的系统：东京大学的研究小组开发了根据文本中出现的表面线索获取单词语义分类的程序。松下研究小组开发了一种类似的技术，使用更深层次的句法结构的文本。这些系统被应用于基因组文本中的文档，有关股票市场的新闻文章等。

The goal of this project was to provide the systems that can acquire knowledge on terminology from texts in a semi-automatic manner. In order to accomplish the goal, we have developed the following three systems.1. Central Database for Terminology : We have created a database system for terminology by integrating the text/lexicon database developed by EDR and the programming language LiLFeS, which was developed at University of Tokyo for easy and flexible treatment linguistic entities By this system, we can perform a systematic maintenance of the knowledge acquired by the following two systems.2. Systems for term recognition : The research group in the NACSIS introduced a statistical metric to identify technical terminology in texts, and built the programs that can recognize terms using this metric. The group in University of Tokyo attacked the same problem in a different perspective, and succeeded in providing a term recognition method based on character n-grams. Those programs are integrated so that they can work as a front end of the database system described in 1.3. Systems for acquiring ontological knowledge on terms : The research group in University of Tokyo developed the programs for obtaining semantic classifications of words according to surface clues appearing in texts. The Matsushita research group developed a similar technique using deeper syntactic structures of texts. Those systems were applied to the documents in Genome texts, the news articles about stock markets and so on.