On September 2, 1999, Information-Base Function TRC Laboratory issued a press release on "Development of text searching software based on ASsociated Keyword Space (ASKS)" as a research result of RWC Project. The press release is outlined here; for a detailed explanation, visit: http://galaxy.trc.rwcp.or.jp/text/galaxy/index.htm
The Real World Computing Partnership has developed software that can search for an article intelligently by using word associations, and has made the software publicly available on the Internet.
As an example of using this software, imagine that you want information about Sendai station on the Tohoku bullet train. You enter the text "Sendai station of Tohoku bullet train" on the screen. This is the input used for the search.
There are four key words of "Tohoku" "bullet train" "Sendai" and "station" as the search input sentence.
When using conventional search software, results are limited since the system searches for documents that include all the key words in a rigid manner (called an "AND" search). Using a conventional method on the Internet database yields only two hits. Next, if we use a conventional method of searching for articles with at least one word matching (called an "OR" search), the number of hits is several hundred.
But by using our newly-developed Association Search Software, if there are few documents that include all the key words, a wider search is implemented by using words that are closely related to each of the key words (associated words) instead of the key words. On the contrary, if there are too many documents that match the key words, the documents that contain more associated words will have higher priority than those containing fewer associated words.
Since we are able to associate from "Tohoku" to "Nagano" in the above example, for example, we can even pick up documents related to "Nagano bullet train" in the search. Since next to "bullet train" we found keywords such as "extending the railroad" and "opening," the search mechanism will prioritize items that include more associated words from the expanded target and from the searched documents.
Also, the words associated with each of the key words can be displayed when searching and the relation expressed with a figure, and so this method automatically presents a new viewpoint in order to obtain new knowledge.
Of course, since this information is generated from an actual database, it includes many items that are "noise". However, we consider that it is able to support us in generating ideas because it displays unexpected words and enables us to search for documents.
The newly-developed "searching software based on association" can also be used in combination with other web search browser such as Yahoo.
Anyone can use "Association Search Software" on the Internet (http://galaxy.trc.rwcp.or.jp/text/galaxy/index.htm).
Users are so far restricted to:
(1) Searches among 2,500,000 submitted news articles, called the FJ category
(2) Searches in combination with a general search browser such as Yahoo
We were able to put our Association Search Software to practical use by developing an automatic method of creating an association dictionary with 450,000 words in this laboratory.
A method of automatically creating this association dictionary is called "Galaxy clustering" and enables automatic allocation of associated words to each word.
Please try using ASKS at the above URL.