中文信息检索中二元文法索引策略的改进

The improved bigram indexing strategy for Chinese Information Retrieval

本文将部分语义信息加入到二元文法中,提出改进的二元文法索引策略。本文应用2-泊松模型的BM25公式在TREC公开数据集上进行了测试。实验表明,改进的二元文法索引策略与基于字的索引策略、基于词的索引策略和基于二元文法的索引策略对比,在主要性能评测参数平均精确率、R-精确率参数上相对较优。

This paper focuses on the indexing strategy for Chinese Information Retrieval（IR）.The improved bigram indexing strategy is put forward by adding semantic information into bigram.2-Possion Model,the classical probabilistic retrieval model,is used as retrieval model.The effectiveness of the new approach is evaluated on TREC Mandarin corpus.Experimental results show that the improved bigram indexing strategy achieves better than the traditional indexing unite,i.e.character,word and bigram,by the measurement of average precision and R-precision.