Rapid porting of the Parlance natural language interface
HLT '89 Proceedings of the workshop on Speech and Natural Language
The BBN Spoken Language System
HLT '89 Proceedings of the workshop on Speech and Natural Language
TEAM: a transportable natural-language interface system
ANLC '83 Proceedings of the first conference on Applied natural language processing
Semantic acquisition in TELI: a transportable, user-customized natural language processor
ACL '86 Proceedings of the 24th annual meeting on Association for Computational Linguistics
A simple statistical class grammar for measuring speech recognition performance
HLT '89 Proceedings of the workshop on Speech and Natural Language
BBN ATIS system progress report—June 1990
HLT '90 Proceedings of the workshop on Speech and Natural Language
(Almost) automatic semantic feature extraction from technical text
HLT '94 Proceedings of the workshop on Human Language Technology
Hi-index | 0.00 |
Acquiring syntactic and semantic information about a new application domain for a natural language processing system is often a time-consuming task. To address this problem, various researchers have developed acquisition tools to speed the process. While such tools are very useful, they are typically tied to particular systems and so their benefits cannot be shared by other researchers.In this paper, we discuss an experiment using the Learner—a software tool for acquiring information about a new task domain for Parlance,1 an ATN-based natural language system—to configure a quite different natural language system, the BBN ACFG, a unification-based system.We have used the Learner to produce information in three major areas: syntactic and semantic information about the lexical items used in the new domain; translation rules from the parser output to the application system; and a class grammar for use in the speech recognition component of HARC, the BBN spoken language system.Initial results are encouraging: 1499 lexical items have been acquired, of which 91% were directly usable, without any manual editing; all of the translation rules are usable; and a speech vocabulary of 2170 items, with an associated class grammar with a perplexity of 89, has been acquired with a small amount of manual editing.