System and method for performing similarity searching using pointer optimization

Abstract

The present invention is a computer-implemented method for optimizing the detecting and scoring of similarities between documents in a source database and search criteria. It uses a set of hierarchical documents having root, interior and leaf nodes where the leaf nodes contain data items. Unique identifiers are assigned to each unique data item contained within each leaf node within the set of documents. Unique identifiers may be assigned to the data items depending upon a selected context in the hierarchy of nodes within the documents. For each data item in each leaf node, a data item score is computed that represents a similarity between the data item in the leaf node and the search criteria. A parent nodes score is then computed by combining the data item scores for all its child nodes. A data and indexing structures provides for efficient similarity searching and the quick reporting of results because the data is organized by the categories a user wants to search. Assigning a pointer or unique identifier only to unique data items causes the memory requirements to be reduced and optimizes the search time. Depending upon the particular set of documents being searched and the search criteria, significant improvements in speed and memory requirements may be achieved.