文本信息检索的代数模型综述

Algebraic Models of Text Retrieval Model: Overview

为了比较各类模型的性能,研究进一步改进文本信息检索各类模型的可能方法,综合已有的试验和研究报告,概述了文本信息检索模型中一类模型——代数模型概念和研究现状,并分析了目前文本信息检索的相关性、相似度、召回率、准确率等概念。结果表明,这几个模型尽管来源于神经网络、向量空间等不同的思想,广义向量空间模型、潜在语义标引模型的计算复杂度比向量空间模型高,但此类模型依据评测在准确率和召回率上相近,经典向量空间模型相对而言,有简单、复杂度低和可作为检索框架,而不仅仅是检索模型的特点。为便于研究者研究、改进、评测信息检索的代数模型,评价了权威的TREC(The Text Retrieval Conference)和两个代表性的研究系统Smart和Lemur。

As the importance of the text retrieval increase, many models have been presented. To compare the performance of models and search for the possible approach to improve the models, the article outlines one kind of IR model the algebraic models and the situations of contemporary research, based on the experiments and research of the domain, analysis＇s relevance, similarity, algebraic model, recall, precision, concluding that the algebraic models are all similar in their recall and precision and the space vector model is simple, low-complexity, and a retrieval frame rather than only a retrieval model although these models are resulted from different ideas. The TREC （Text Retrieval Conference） and some software for experiment and research are introduced and evaluated briefly for researcher to study and develop the algebraic model.