As the methodology of Open Complex Giant Systems (OCGSs), meta-synthesis was proposed in the early 1990s by Prof. H. S. Tsien and et al. Its central principle is to combine the intelligence of human and the computational efficiency of computers, aiming to handle the problems those cannot be solved by human or computer alone. In 1992, meta-synthesis was developed to HWME (Hall for Workshop of Meta-synthetic Engineering), which emphasized the collective wisdom. The collective wisdom is derived from discussions of human experts and cooperation between human and computers. Authoritative information relevant to discussion topics has showed its great importance in HWME applications. And it's clear that the Internet is a good resource for such information. However, in HWME applications, human experts often work under severe time pressure. It's hard for them to search information on the internet frequently and timely. Thus, an active information retrieval method is essential for HWME applications, which can automatically sense topics, generate query terms and search on the Internet. This dissertation mainly focuses on developing such an efficient active information retrieval method. And the main achievements of our research are listed as follows: 1. An active information retrieval framework for HWME was proposed. The framework consists of four basic modules: a domain thesaurus construction module (module A), a discussion texts analysis module (module B), an information retrieval and filtering module (module C), a personalized information recommendation module (module D). Before discussions, structural multi-domains thesauri are constructed based on the background materials relevant to the discussions by module A. During a discussion process, the topics are extracted and the change of topics is judged. And query terms are generated and sent to a search engine by module B. The retrieval results are obtained and filtered according to the discussions by module C. The information that users are interested in is selected and personalized information is recommended to the right users by module D. The framework is proved to be suitable for HWME applications, which often has the following characteristics: the discussions often have a strong domain background, while the topics changes frequently. 2. An automatic method for building structural domain-specific thesauri is proposed. This method consists of two steps: a) discovering domain-specific terms from document co...
修改评论