NATURAL LANGUAGE INTERFACE FOR MULTI-AGENT CONTRACTING SYSTEM (MACS)

摘要

Facing serious imbalances in the skills and experience of our highly talented and specialized civilian workforce, the Naval Postgraduate School has initiated the External Acquisition Research Program (EARP) to promote advances in basic and applied research as related to US defense acquisition. One of the projects funded under EARP over a 2 year period is the Multi-Agent Contracting System (MACS) project (Liebowitz et al., 2000. Journal of Knowledge-Based Systems 13: 241-250). MACS was designed to provide advice in the pre-award phase of a defense contract. The original version of MACS allowed users to query the system using a list of predetermined from pull-down menus. An exact string of had to be used to obtain an answer from MACS, forcing the users to memorize a list of keywords. This inflexible user interface could have caused problems for those users who did not know the right and, in turn, reduced the usability of MACS. The objective of this study is to extend the capability of MACS by allowing users to interface with the system using natural language. The natural language user interface will provide the users with more flexibility in formulating queries on the pre-award contracts, increasing the user friendliness of its interface. The paper presents the methods used to develop a natural language user interface for MACS.