Statistical Phrase Extraction Techniques In Natural Language Databases.

在自然语言文本中定位重要短语的能力对于在文本中建立索引或放置超链接非常有用。在任何一种情况下，人们都力求改善对文本材料的获取。在过去，最常用的短语定位方法是词性标注。我们开发了一种新的方法，使用评分算法对短语的有用程度进行排名。已经开发和测试了许多不同的方法。为了检索的目的，这些正在与词干提取和寻找同义短语的屈折变体的方法相结合。UMLS系统也被用来寻找同义词短语的索引。这些方法被用于在NCBI的电子教科书项目中寻找有用的短语，该项目目前已上线，但仍在开发中。这些方法也开始应用于PubMedCentral的生物学和医学期刊文章数据库，以及对该数据库中过期期刊扫描的OCR材料进行索引。目前的工作重点是为该系统增加一个复杂的缩写检测能力。

The ability to locate important phrases in natural language text is useful for the purposes of indexing or placing hyperlinks in text. In either case one seeks to improve access to the textual material. In the past the most common method used for the location of phrases has been a part of speech tagger. We have developed a new approach that uses scoring algorithms to rank phrases as to how useful they may be. A number of different methods have been developed and tested. These are being combined with methods of stemming and of finding inflectional variants of phrases that are synonymous for retrieval purposes. The UMLS system is also being used to find synonymous phrases for indexing. These methods are being applied to find useful phrases in NCBI's electronic textbook project that is currently online but still under development. The methods are also beginning to be applied to the PubMedCentral database of journal articles in biology and medicine and to the indexing of OCR material from the scanning of back issues of journals for this database. Current work is focused on adding a sophisticated abbreviation detection capability to this system.