Desktop search system based on the Action-Oriented algorism

摘要

Have you ever been unable to retrieve a web page even though you have read it before? In Japan, the broadband environment is common now, and people easily read many pages obtained through the Internet. However, finding the files is sometimes difficult because of forgetting how to find them. Some desktop search services have begun to be provided to solve this problem. However, users do not seem to be satisfied with them. For example, most services show many unimportant files. Against this background, we propose a new desktop search system that uses text of the files and various action logs of a user such as view time, view period, frequency of copying the file, and frequency of printing the file. The system shows files that are more useful from among many files that have been read. In this paper, the outline of the desktop search prototype system and the result of an evaluation experiment are described. To collect various action logs without user effort, we developed a logging agent that gathers and stores the logs automatically. Action-Oriented algorithm deduces the degree of importance of a file for each person by analyzing the various action logs. The proposed system shows thumbnails of files that are prized highly valuable according to their degree of importance and the result of keyword matching. In the evaluation experiment, we assume that the service's goal is to enable retrieving of an important page within a short time.