Intelligent Agent for Matching Occupations, Personnel and Trng Materials

Latent Semantic Analysis (LSA) is a machine learning method that extracts contextual meaning similarities among words and passages by analysis of large bodies of natural text. We will test the feasibility and effectiveness of incorporating LSA into a web-based search agent that can compare the conceptual content of: (a) textual training materials, (b) descriptions of personnel competency requirements, (c) descriptions of civilian occupations, (d) descriptions of individual training, experience, or test performance. The experimental system will provide platform-independent access to a multimodal web-pager interface for entering descriptions and displaying relevance ranked results. As proofs-of-concept, we will use the agent to: (a) identify and rank the whole and each paragraph of principal textual materials for at least 30 AF courses according to the relevance of their content to the competencies required by a selected military system, and (b) produce a ranked list of the conceptual similarity of each of at least 30 AF occupations to the 20 most similar civilian occupations described in the Department of Labor Occupational Network. We will assess the validity of the LSA measure in both applications by comparison with judgments by subject matter experts with respect to a sample of cases.