We are employing JICST magnetic tapes either for retrieving documents relevant to our own research and for using them as experimental data necessary to develop document systems with high freedom based on relational database model and various techniques for processing of bibliographic information. In this report the results so far obtained are summarized. First, studies on relational databases and files, and development of laboratory information system are outlined together with those studies as reported here. Then, the performance of JICST tape retrieval system in our laboratory follows. The major topic ofthe present paper lies in the newly developed techniques such as RS Index, MULTI-KWIC, multi-level partial matching, etc., and they are illustrated with examples. The RS Index system, i.e., Reference Structure Index, can output reference structures in tree form, and help finding out related bibliographic references or the research trends. It is characterized byan algorithm of small memory capacity using pushdown stack, and to its basic algorithm various functions are added to form the program. Hitherto widely used KWIC is generalizedto make up a multi-functional MULTI-KWIC. This system has a key-phrase extracting function, which is developed with particular attention to the set of document titles, and cangive a KWIC index of easily understandable key phrases, a KWIC system consisting of a pair of keywords (particularly called MULTI-KWIC Index), and annual use frequencies of keyphrases, etc. Finally, a multi-level partial matching algorithm as an efficient technique for partial matching of strings is described. This method reduces computing time bypre-processing, and an experiment using JICST tapes shows a further reduction of comparison frequencies of the Aho's method, which is known to be the best, to 1/2-1/5.
View full abstract