A Contrast Between Computer and Human Language Understanding
Abstract
This paper is a progress report of the natural language (NL) understanding portion of the Artificial Intelligence Measurement System (AIMS). The focus of the document is on the research plan and initial steps in relating human performance measures to NL implementations. The overall goal was to norm a given NL system's performance on a sample of people. To this end a test was created to capture the language functions of a target implementation but was set within the domain of the human population to be tested. The NL system used was IRUS (Bates, Stallard, & Moser, 1985), a syntactic shell interface available from a system test in a military environment. A syntactic analysis of the language IRUS could handle suggested that an initial human population for comparison should be very young children. This paper provides the first steps in the development of a test instrument that contains the syntactic structures of IRUS set within a very simple database, one consisting of animals, people, houses, their attributes and positions.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1988
- Accession Number
- ADA232917
Entities
People
- Elaine L. Lindheim
- Eva L. Baker
Organizations
- University of California, Los Angeles