Methods of evaluating semantic differences, methods of identifying related sets of items in semantic spaces, and systems and computer program products for implementing the same

摘要

One aspect of the invention provides a method of evaluating semantic differences between a first item in a first semantic space and a second item in a second semantic space. The method includes: calculating a first ordered list of N nearest neighbors of the first item within the first semantic space; calculating a second ordered list of N nearest neighbors of the second item within the second semantic space; and computing a plurality of similarity measures between the first n nearest neighbors of the first item and the first n nearest neighbors of the second item, wherein n and N are positive integers and 1≦n≦N.

主权项

1. A method of identifying a first set of items from a first semantic space that are semantically related to a second set of items from a second semantic space, the method comprising:
generating an induced semantic space for the first semantic space, the induced semantic space represented as a square matrix in which each matrix entry (i, j) is a numeric representation of an association strength between item i and item j in the first set of items; setting matrix entries below a defined replacement threshold to 0; selecting e items having the highest value entries from the j columns; ranking the e items by association strength entry values; selecting n items having the highest value entries from the e items; calculating a similarity measure between each of the n items selected from the first semantic space and each of the n items selected from the second semantic space; and calculating an aggregate similarity value from the similarity measures between each of the n items selected from the first semantic space and each of the n items selected from the second semantic space.