SDW'98 $BH/I=O@J8(B #47 ( $BEE;R=PHG$N%Z!<%8$X(B )
Importance of digital documents is growing with the recent advances of computing environments. The volume of document data is huge and it is not easy to access relevent information. In this paper, we propose a method to extract logical structures embeded in documents and retrieve relevent passages from documents flexibly. This method makes it possible to use passages of various abstraction levels as units in document retrieval. Experimental evaluation of this method is given. We also describe a technique to dynamically and locally extract logical strctures using inverted file strucures.