Portability in the Janus Natural Language Interface

Although natural language technology has achieved a high degree of domain independence through separating domain-independent modules from domain-dependent knowledge bases, portability, as measured by effort to move from one application to another, is still a problem. Here we describe a knowledge acquisition tool (KNACQ) that has sharply decreased our effort in building knowledge bases. The knowledge bases acquired with KNACQ are used by both the understanding components and the generation components of Janus. INTRODUCTION: MOTIVATION Portability is measurable by the person-effort expended to achieve a pre-specified degree of coverage, given an application program. Factoring an NL system into domain-dependent and domain-independent modules is now part of the state of the art; therefore, the challenge in portability is reducing the effort needed to create domain-dependent modules. For us, those are the domain-dependent knowledge bases, e.g., lexical syntax, lexical semantics, domain models, and transformations specific to the target application system. Our experience in installing our natural language interface as part of DARPA's Fleet Command Center Battle Management Program (FCCBMP) iUustrates the kind of portability needed if NL applications (or products) are to become widespread. We demonstrated broad linguistic coverage across 40 fields of a large Oracle database, the Integrated Data Base (IDB), in August 1986. A conclusion was that the state of the art in understanding was adequate. However, the time and cost needed to cover all 400 fields of the IDB in 1986 and the more than 850 fields today would have been prohibitive without a breakthrough in knowledge acquisition and maintenance tools. We have developed a suite of tools to greatly increase our productivity in porting BBN's Janus NL understanding and generation system to new domains. KREME [Abrett, 1987] enables creating, browsing, and maintaining of taxonomic knowledge bases. IRACQ [Ayuso, 1987] supports learning lexical semantics from examples with only one unknown word. Both of those tools were used in preparing the FCCBMP demonstratior~ in 1986. What was missing was a way to rapidly infer the knowledge bases for the overwhelming majority of words used in accessing fields. Then one could bootstrap using IRACQ to acquire more complex lexical items. We have developed and used such a tool called KNACQ (for KNowledge ACQuisition). The efficiency we have experienced results from (1) identifying regularities in expression corresponding to domain model structures and (2) requiring little information from the user to identi~ expressions corresponding to those regularities. 1 This research was supported by the Advanced Research Projects Agency of the Department of Defense and was monitored by ONR under Contracts N00014-85-C-0079 and N00014-85-C-0016. The views and conclusions contained in this document are those of the authors and should not be interpreted as necessarily representing the official policies, either expressed or implied, of the Defense Advanced Research Projects Agency or the U.S. Government.