The Word-Based Pyramid Method

Report No. ARL-TR-5912
Authors: Andrew A. Thompson
Date/Pages: February 2012; 24 pages
Abstract: The comparison of human and machine generated descriptions is a problem confronting the document-understanding community and other researchers developing systems that output natural language. An investigator often wants to know how similar a machine description is to a human description. By assuming the message can be portrayed by a "bag of words",it is possible to automate model formulation and then evaluate descriptions. Based on the study performed, the word-based pyramid method provides a useful tool for the quantification of verbal similarity.
Distribution: Approved for public release
  Download Report ( 0.229 MBytes )
If you are visually impaired or need a physical copy of this report, please visit and contact DTIC.

Last Update / Reviewed: February 1, 2012