Semantic-Based Feature Vector Matching for Heterogeneous Databases

Searching for special information in today's information explosion becomes more and more difficult. Using large search engines like Google is not suitable because they usually find millions of results out of millions of databases. In this paper, we investigated a new method to avoid the user's trouble of find what they need from massive data found by global engines. A new search framework based on semantic ontology and data mining is proposed to find most suitable and valuable results, according to personal attributes, from selected heterogeneous databases. Within this framework, we use query vector to describe the search goal, user vector to describe attributes of user, database vector to describe attributes of databases and case vector to describe features of a case. After a match of user vector and databases vector, the databases with most match attributes are selected. Then a distance between query vector and case vector is computed, influenced by the match level of user vector and databases vector. And a method "Vectors Match" to get and sort the best results by the distance between a case and the query. Desktop software based on this framework is proved that can provide more meaningful and exhaustive query results.