事件信息抽取中的数据预处理方法研究

摘要

Event extraction is an important area in information extraction research. Due to such problems as incomplete information, unclear semanteme, diversified elementary expression and obvious event redundancy with event extraction, the thesis proposes both missing data filling algorithm based on statistics to perfect events with missing information, and event element standardisation based on rules and dictionaries to unify events which are expressed differently. By authenticating events it solves the problem of semantic ambiguity, fixes incorrect event extraction, at the mean time filters out events with obvious redundant information.%事件抽取是信息抽取领域的一个重要研究方向.针对事件抽取获得的信息不完整、语义不明确、元素表达多样性及明显事件冗余等问题,提出基于统计的缺失数据填充算法,使丢失信患的事件完备化;同时提出基于规则和词典的事件元素规格化将不同表述的事件统一化,通过事件真伪辨别解决了语义不明确问题,修正抽取不正确的事件,并过滤掉明显冗余信息的事件.