Finding effective query strings from results of primary search

This paper proposes a method to find query strings suitable for successive searches from primary search results. This method may be regarded as a novel kind of keyword extraction for information retrieval, where these strings are extracted from primary search results. These strings are selected depending on the following conditions: effectiveness, prevalence, and uniqueness. In addition, this method does not use any kind of dictionary, not even a Japanese morphological analyzer. The proposed procedure consists of two parts. The first part is selecting the first candidates, which are all of the keywords in primary search results. The second part is narrowing down the candidates so that the candidates form reasonable cluster of primary search results. Our main concern is whether the selected string is meaningful or understandable for people. We have found that more than 90% of the strings that satisfy the conditions above are meaningful and correct Japanese words.