Overview of the Core Language Engine 1 Design Goals and Architecture 2 Category and Feature System

The Core Language Engine (CLE) is a domain independent system for translating natural language (English) sentences into formal representations of their literal meanings which are capable of supporting reasoning. It is designed to be used as a major component of interactive advisor systems such as interfaces to database management systems and diagnostic expert systems. The main contribution of the CLE is intended to be substantial coverage of English constructions in both syntax and semantics that is well motivated and hence extensible. Interactive facilities are provided to allow users to extend the system vocabulary. The CLE has a modular architecture with well deened interfaces between the various stages of linguistic processing. Uniication is used as the mechanism for rule application and passing information in each of morphology, parsing, interpretation, and selectional ltering, so the rules for these components are expressed declaratively. A compact representation of local ambiguities is applied systematically in syntax and semantics, allowing us to adopt the modular staged approach without sacriicing computational eeciency. The CLE is a domain independent system for natural language processing which can be used as the basis for building natural language applications. It is primarily aimed 1 at interactive advisor systems such as interfaces to database management systems and diagnostic expert systems. English sentences are translated by the CLE into logical form expressions which represent their literal meaning. The motivation for choosing such representations as the output of the system is that they are capable of supporting reasoning which is necessary both for natural language disambiguation and for the application task of typical advisor systems. The main contribution of the CLE is intended to be substantial coverage of in both syntax and semantics. This coverage is achieved by taking advantage of recent advances in linguistics giving analyses that are well motivated and hence extensi-ble. Transportability to new applications is also enhanced by providing interactive facilities to allow users to extend the system vocabulary. Current CLE coverage of English syntax and semantics can be summarized roughly as follows: Noun phrases: prenominal and postnominal modiiers, lexical and phrasal quantiiers/speciiers. Coordination: conjunctions and disjunctions of a wide class of noun phrases, verb phrases, and clauses; some common comparatives. Morphology: innectional morphology, simple productive cases of derivational morphology. Core lexicon: 1200 function words and content word stems, 1800 senses and their selectional restrictions. The CLE has a modular architecture. Sentence processing is performed in the following stages: { …