基于单字特征和搜索引擎的新词识别

发布时间：2018-04-04 13:26

本文选题：新词识别　切入点：单字词　出处：《武汉大学学报(理学版)》2010年06期

【摘要】：新词识别是影响搜索准确率以及速率的重要因素.本文提出了一种基于统计模型和词语搭配的中文新词自动识别方法.采用条件概率的方法提取单字词搭配特征和临界词特征,并采用层次结构实现新词定位以及识别.首先采用双向最大匹配相结合的方法对文本进行词法粗切分,然后根据单字词搭配得到候选新词的位置,用临界词方法确定候选新词的边界,采用改进Nagao串频统计方法对新词候选词在本文内进行重复串统计,对于只在文中出现一次的新词则借助搜索引擎进行确定.对新浪网近期的网络文章进行测试,结果表明,基于本文方法设计的系统可以识别不同领域的新词,在低频词、较长的词以及新词语搭配方面取得了良好的效果.单字词搭配检查发现新词位置综合指标F值达到96.8%.
[Abstract]:Neologism recognition is an important factor affecting search accuracy and speed.This paper presents a new Chinese word recognition method based on statistical model and word collocation.The method of conditional probability is used to extract the collocation feature and critical word feature of single word, and the hierarchical structure is used to locate and recognize new words.First of all, the text is divided into lexical coarse segments by using bidirectional maximum matching method, and then the boundary of candidate neologisms is determined by critical word method according to the location of candidate neologisms according to the collocation of single word words.The improved Nagao string frequency statistics method is used to make repeated string statistics of new word candidates in this paper, and the search engine is used to determine the new words that appear only once in this paper.The results show that the system based on this method can identify new words in different fields and has good results in low frequency words, long words and collocation.Single word collocation check found that the comprehensive index F value of neologism reached 96.8%.
【作者单位】：武汉大学计算机学院;
【分类号】：TP391.1

【参考文献】