Natural language project: Istituto di Psicologia, Roma, Italy
暂无分享,去创建一个
(1) Explicit and detailed specification of performance to be obtained and of input to system. (2) "Occam's Razor: Do not introduce anything on the basis of tradition or a priori philosophy (e.g. NP, [for all] x, extension vs. intension). Introduce only such structure as is necessary to produce the performance to be obtained. (3) Do not restrict excessively the range of linguistic facts to be considered. The system must comprehend or produce at least 75% of a page in a newspaper. (4) Do not try to obtain directly overall and "deep" comrehension. First get literal comprehension, close to linguistic input. Obtain the rest with successive processes. (5) Avoid "cosmetics": e.g. in production don't have the machine produce expressions that "ape" human behavior but that the machine does not understand. (6) Stay close to natural logic: vagueness, contradictions, natural language quantifiers, not absolute and not metric time, etc.
[1] Domenico Parisi,et al. A goal analysis of some pragmatic aspects of language , 1981 .
[2] Oliviero Stock,et al. Strategy Selection for an ATN Syntactic Parser , 1980, ACL.
[3] Norman E. Eliason,et al. Essentials of Grammar , 1941 .
[4] Oliviero Stock,et al. Automatic analysis of Italian , 1980 .