UNL System

Revision as of 15:18, 24 September 2012

The UNL System is the set of basic modules of UNL and the rules to combine them.

UNLization

Main article: UNLization

UNLization is the process of representing the content of a natural language structure into UNL. This process is carried out by a UNLizer. The input for the UNLization is a natural language document or set of documents. The input document is first segmented, according to the boundaries defined in the NL-UNL T-Grammar. Next, each processing unit is tokenized, according to the NL-UNL Dictionary and the NL-UNL Memory. The tokenization process may be controlled by the NL-UNL D-Grammar. The resulting tokenized string is then syntactically and semantically analyzed with the NL-UNL T-Grammar, which may be improved by the NL-UNL D-Grammar, the NL Memory, the UNL Memory, the UNL Knowledge Base and the UNL Ontology. The output of the process is a UNL document, which may reflect either the deep or the surface structure of the text, of each sentence or of each word of the natural language input document, depending on the structure of the grammar.

NLization

Main article: NLization

NLization is the process of generating a natural language document out of UNL. The input for the NLization is a UNL document. The input document, which is already segmented according to the UNL Document Structure, is [tokenization|tokenized]], according to the UNL-NL Dictionary and the UNL-NL Memory. The tokenization process may be controlled by the UNL-NL D-Grammar. The resulting tokenized string is then syntactically and semantically analyzed with the UNL-NL T-Grammar, which may be improved by the UNL-NL D-Grammar, the NL Memory, the UNL Memory, the UNL Knowledge Base and the UNL Ontology. The output of the process is a natural language document, which may reflect either the deep or the surface structure of the whole or part of the UNL input document.

Modules

• Lexical Databases
  • The UNL Ontology, the UNL Knowledge Base and the UNL Memory are lists of semantic frames between UW's. These repositories include ontological relations ("a kind of" and "an instance of"), in case of UNL Ontology; necessary (ontological and thematic) relations (such as "is the agent of", "is the place where", "is the moment when"), in case of UNL Knowledge Base; and any typical relations, in case of the UNL Memory.
  • The UNL-NL Dictionary and the NL-UNL Dictionary contain mappings between UW's and natural language entries. They are bilingual lexica where UW's are translated into lexical realisation units of a given natural language.
  • The UNL-NL Memory and the NL-UNL Memory contain mappings between UNL and natural language entries.
  • The NL Memory contains syntactic relations between natural language entries
• Grammars
  • The UNL-NL D-Grammar and the UNL-NL T-Grammar are the set of rules used for natural language generation (out of UNL). The NL-UNL D-Grammar and the NL-UNL T-Grammar are the set of rules used for natural language analysis (to UNL). The T-Grammar is the repository of the transformation rules, which are used to manipulate data structures; the D-Grammar is the repository of disambiguation rules, which are used to control the use of transformation rules.

Specs

Tagset

The Tagset is a harmonized system for representing linguistic constants.

Dictionary Specs

The Dictionary Specs state the syntax of dictionaries.

Grammar Specs

The Grammar Specs state the syntax of grammar rules.

KB Specs

The UNL Knowledge Base Specs state the syntax of knowledge bases.

Memory Specs

The Memory Specs state the syntax of example bases.