A Knowledge-Based Approach for Searching Semi-Structured Documents

The amount of information on the Internet is increasing dramatically, making efficient Information Retrieval (IR) very difficult. Most of the search engines are using traditional IR methods based on keyword matching and are already showing a number of limitations, particularly they suffer from low precision. One suggestion for improving their performance is to use Natural Language Processing (NLP) technology to support content-based IR. However, current NLP technology is not mature enough to enable free text parsing and understanding. This research focuses on searching semi-structured documents, such as Web pages generated from databases and online services. Shallow NLP and a knowledgebased approach is used to parse both users' query and semi-structured documents to a set of knowledge units (concepts) and the knowledge units (concepts) are used as the basis for matching instead of keywords. An online Classified Advertisement Search Agent (CASA) has been built as an example. It can automatically search through online real estate advertisements to help users find accommodation. This paper focuses on the knowledge unit representation and text parsing algorithm which are used for parsing semi-structured documents and users' query