ABSTRACT
Machine Translation (MT) technology has progressed significantly since the days of the ALPAC* report. In particular, multiple paradigms are being investigated ranging from statistical methods to full knowledge-based interlingual MT systems. Much of the recent work is based on advances in natural language processing since ALPAC in the 1960s, including:• Semantic analysis to resolve lexical and syntactic ambiguities during parsing, and thus reduce translation errors very significantly.• Unification grammars allowing syntactic and semantic constraints to be checked in a unified manner while parsing, and permitting reversible grammars---i. e., the same grammars to be used for generation as well as for analysis.• Advanced parsing methodologies, including augmented-LR compilation where knowledge sources (syntactic grammars, lexicons, and semantic ontologies) can be defined and maintained separately but are jointly compiled to apply simultaneously at run time, both in parsing and in generation.• Natural language generation, focusing on how to structure fluent target-language output, an activity not truly investigated in the pre-ALPAC days.• Automated corpus analysis tools, statistical and other means of extracting useful information from large bi- or multi-lingual corpora, including collocations, transfers, and contextual cues for disambiguation.• MRDs → MTDs, use of electronic machine-readable dictionaries (MRDs) to partially automate the creation of machine-tractable dictionaries (MTDs) in processable internal form for parsers and generators, permitting principled scaling up in MT configurations.
- Session summary
Recommendations
Session summary
HLT '90: Proceedings of the workshop on Speech and Natural LanguageThe papers in this session addressed issues in combining speech recognition with natural language systems. The first three papers concern the use of grammars. Speech recognizers and Natural Language parsers make different requirements of language ...
Session summary
HLT '91: Proceedings of the workshop on Speech and Natural LanguageThe first natural language session concentrates on issues of parsing sentences into representation structure. Roughly, there are two parts to this problem:• finding a description of the structure of a natural language such as English, namely specifying ...
Session summary
HLT '93: Proceedings of the workshop on Human Language TechnologyCollectively, the papers in this session are mainly concerned with parsing, semantic interpretation, and inference. Interest in these processes can be motivated if we recognize that the overall goal of the field of NLP is the manipulation of natural ...
Comments