The Research And Implementation Of Full-Text Retrieval Based On Byte Stream

Full-Text Retrieval is one of the important aspects in current information retrieval technology. It is an effective way of obtaining useful information in massive document warehouses. The key issue of implementing a full-text retrieval system is building an effective model. With respect to the index item, traditional full-text retrieval models can be divided into keyword-based models and character-based models. Compared with character-based models, keyword-based methods are faster, with lower space cost and easy to support synonym queries. Hence various commercial full-text retrieval methods are based on keyword-based models. However, keyword-based methods are seriously challenged by the following problems: 1) its recall and precision are relatively low; 2) it does not support full-sentence query and document fraction query; 3) it can not handle Chinese effectively, especially in word syncopation.In the passed few years, character-based methods are widely studied for the ability of solving above problems with reasonable speed, but still lower than keywords-based methods. In this paper, character-based methods went by the name of byte stream based methods in order to apply them in all kinds of files.To solve above problems, this thesis makes following contributions:(1) Analyzes the basic methods and common models of full-text retrieval technology, and does an in-depth research on byte stream based full-text retrieval.(2) Proposes three full-text retrieval models based on byte stream: Cross Linked Model (CLM), Inversed Subsequence Model (ISM) and Adjacent Address...