Development of Understanding System for Scientific and Technical Documents

该项目的目标是开发一个科技文件理解系统，如INSPEC数据库。这一目标要求我们开发将句子转换为一阶公式的技术，其中谓词符号是动词，术语是名词短语。因此，我们研究了动词短语结构的确定和名词短语范围的确定问题，并取得了以下成果：(1)提出了一种基于摘要中单词的出现频率与关键短语的比例来确定动词短语的方法。(2)名词短语可以被认为是由简单的名词短语组成的，每个名词短语都由一个中心名词和它的介词修饰语组成。提出了中心名词和介词修饰语的识别方法。(3)传统的识别方法是利用可能包含词汇歧义的词性信息进行识别。实验表明，(2)的方法优于传统的方法。(4)利用(2)的识别方法，我们提出了一种简单名词短语范围的标引方法。

The goal of this project is to develop an understanding system for scientific and technical documents, such as INSPEC database. This goal requires us to develop techniquies of converting sentences into the first order formulas where the predicate symbols are verbs and the terms are noun phrases. We therefore addressed the problems of determination of structure of verb phrases and identification of ranges of noun phrases, and achieved the following resuls.(1) We developed a method of determining the verb phrases which is based on the ratio of the frequencies of a word in a abstract and key phrases.(2) Noun phrases can be considered to be composed of simple noun phrases, each of which consists of a head noun and its prepositional modifiers. We developed method of identifying head nouns and prepositional modifiers.(3) Traditionally the above mentioned identification was carried out by using parts-of-speech information which may contain lexical ambiguities. We showed the method of (2) is better than the traditional method by experments.(4) Using the identification method of (2) we developed a method of indetifying the ranges of simple noun phrases.