数值信息抽取研究进展综述

Numerical Information Extraction: A Review of Research

通过对数值信息抽取文献的调研,先从文献类型、学科领域、高频关键词三个方面进行定量分析,从抽取数据源、抽取对象、抽取方法与技术、结果评价和应用等方面对当前数值信息抽取研究进行了梳理和总结。研究发现当前对于数值信息抽取的研究具有五个特点:抽取数据源以新闻语料、Web网页为主,抽取对象以基数类数值信息和数量类数值信息为主,抽取方法以基于规则的方法为主,抽取结果评价指标比较单一,但应用领域较为广泛。图4。表3。参考文献56。

This paper first makes a quantitative analysis on the documents of numerical information extraction from three as- pects： document type, subject area and high frequency keywords. Then the research context is summarized from four aspects： data source type, object for extraction, extraction method and technique, result evaluation and corresponding application. Our findings are as follows： news corpus and web pages are the main data sources; cardinal numbers and quantitative phrases are the main objects for extraction; extraction method and technique are mainly rule-based and the result evaluation indicators are rela- tively simple but have a wide scope for application. 4 figs. 3 tabs. 56 refs.