Tomita, Masaru Carbonell, Jaime G. The Universal Parser Architecture for Knowledge-Based Machine Translation <p>Machine translation should be semanticalty-accurate, linguistically-principled, user-interactive, and extensible to multiple languages and domains. This paper presents the<em> universal parser</em> architecture that strives to meet these objectives. In essence, linguistic knowledge bases (syntactic, semantic, lexical, pragmatic), encoded in theoretically-motivated formalisms such as lexical-functional grammars, are unified and precompiled into fast run-time grammars for parsing and generation. Thus, the universal<em> parser</em> provides principled run-time integration of syntax and semantics, while preserving the generality of domain-independent syntactic grammars, and language-independent domain knowledge bases; the optimized cross product is generated automatically in the precornpllation phase. Initial results for bi-directional English-Japanese translation show considerable promise both in terms of demonstrating the theoretical feasibillty of the approach and in terms of subsequent practical utility.</p> Software Research 1987-01-01
    https://kilthub.cmu.edu/articles/journal_contribution/The_Universal_Parser_Architecture_for_Knowledge-Based_Machine_Translation/6626213
10.1184/R1/6626213.v1