Knowledge-Pattern Based Information Extraction

Information Extraction (IE) is a technology for "reading" reports and picking out the bits of information that are needed by users. Hypermedia relies on a combination of knowledge representation e.g. semantic links, text analysis; and "canned" knowledge in different presentation formats. Successful information access and presentation depends on an information base where the information is represented, and not only contains a presentation of the knowledge. A weak representation of the knowledge, and the limitations of the knowledge, lead to difficulties in finding the relevant information, and may also cause the system to retrieve information that is incorrect in the current context, but would be correct in another usage situation. Combining knowledge structures, with "canned" knowledge provides better knowledge structure. This paper presents a proposed framework of how to check and compare the extracted information from different documents and to what extent they are relevant to the user profile. The framework is based on the pattern similarity between the indices of the knowledge representation structure, or the knowledge-patterns, which includes all the elements of the knowledge structure: entities, attributes, actions, scripts, relations, and operators or rules, in addition to the relevant linguistic rules.