| Accompanying with the information resources Internetilation and digitalization, network information resources from quantity to contents have made a breakthrough, presenting characteristic of various types, multi-medias, abnormity, across time, across the geography and across the language and so on. The information resources seems to be the dispersive, out-of-order, change greatly stochastically, presenting complicate content characteristic, its store form is text-hypertext-multimedia-hypermedia, making method of information organization change greatly not only regarding knowledge and information as basic unit, but showing the logic relationship among them, providing technique support for administration and exploiting of information resource under network, through make use of storage technique of digitalization developing from traditional tactic and linear into hypertext, hypermedia technique, making information compose mutually relating ,direct, nonlinear network according the relationship of itself.The characters above bring many obstacles to users and put forward to new problem for structure organization and retrieval strategy of Internet information. To Make full use of network information, we must store and organize information using the scientific, reasonable method according with human thought way.System taxonomy is a grade concept caption system of embodying knowledge classification. It comes into being by carving up and arranging systemically concepts summing up document content and external character. Its main character is integrating documents to show the distinction and relation of various documents in contents. This research studies and discusses how to use knowledge excavation making subject knowledge classification toward the information scattering in database and Internet, exploiting system of knowledge classification and retrieval, taking biomedical engineering for example. The system means to provide a methodfor the building of other subject classification system. In addition, word lists guiding database was established, which makes the use-surface more clarified.The study method: TRS (Full text retrieval system) is selected as Mining tool. Two aspects of this study, one is collecting manually the documents that already exist in database such as CBMdisc, MEDLINE, El, INSPEC, and integrating them into database, providing the function of intelligent retrieval, the other is collecting the Internet html texts automatically into TRS database and issue it. The steps: firstly, set the catalog of classification, according with the 4th version we classifies the BME into the following 4 classes: general problem (including Biomaterial; Biomechanics; Bioinformation); Artificial organ of various systems; other branches (light, laser biomedicine and hypothermia biomedicine) and instrument and equipment. We investigate the embodying and disposal of the main domestic and foreign databases ,select the data that can be used directly , build database and issue it through web application server, which could provide the function of intelligent retrieval. The second aspect: using the intelligent classification training machine of InfoRadar, we provide a group of training texts for every classes. Through automatic learning process, Classification models corresponding with every class come into being. Then News-Robot could collect html information automatically. Finally, Web issue model could issue the text information and provide the retrieval function based on content.System evaluation: the first retrieval system implement the automatization partly from the angle of data mining. According to retrieval result pertinence taxis, selecting the records we need is still a problem, for we could not find a suitable threshold.The second system incarnate the automatization, but the types of the content excavated from Internet is irregular: including the introduction of product, and use manual, conference, and so on. We could make a subdivision further using the same metho...
|
PDF Full Text Request