| | Would you need rich information of the grammatical features of your textual content? Are you designing an application where you would need syntax information? Lingsoft's morphological analyzers can provide all possible grammatical readings of a particular word form. If multiple readings are found, the analyzer cannot judge which of them don't match that particular context. Is there a tool that can? Yes, there is.
The constraint grammar concept was launched by Lingsoft's co-founder Fred Karlsson in 1990, and has since become one of the most renowned techniques in natural language processing. Lingsoft uses this powerful framework for its grammar checkers, regarded as the best of their kind.
A constrain grammar parser needs a morphological analyzer as a subcomponent to provide tagged input. The parser works best with correctly spelled and punctuated standard written language without grammatical errors. Considerable tolerance is however shown for grammatically incorrect language.
Lingsoft's constraint grammar parsers take the morphologically tagged input text, inspect it sentence by sentence and try to remove unlikely or select likely morphosyntactic readings given for the individual words by applying constraint grammar rules. These rules can be based on the word itself, the whole sentence or anything in between. Multiple readings may remain in the output if the parser's set of rules cannot disambiguate between them. The morphopsyntactic tags produced by the morphological analyzer remain in the output, so you can filter, label and organize the output in a desired manner.
The disambiguation rules along with other data, such as the corresponding morphological analyzer, are included in a language-specific constraint grammar module file. As a rule-of-thumb the disambiguation speed of a constraint grammar module is one-tenth of the corresponding morphological analyzer's.
Lingsoft's constraint grammar parser modules are used with Lingsoft's proprietary LSINDEX application programming interface (API), available for Windows, Linux, Mac, Java, and others on demand. The character set used with LSINDEX is Unicode Thus the modules fit almost any integration scenario.
Product specifications | How to buy
| | 
Copyright ©1986-2018, Lingsoft Ltd.
|
