LSS

From UNL Wiki
Jump to: navigation, search

Linear Sentence Structures (LSS) are sequences of part-of-speech tags.

Contents

Methodology

LSS's are extracted from the tokenization of a corpus using the enumerative dictionary, i.e., the list of the word forms available for a given language. As the enumerative dictionary may contain multiword expressions, the length of the LSS may not correspond to the number of isolated words in the sentence, but to the number of dictionary entries resulting from the application of the principle of the longest first. Additionally, as the tokenization does not perform any lexical disambiguation, the LSS contain all possible lexical categories of each of the components. The punctuation of the original sentence is preserved.

Categories

LSS uses the values of the attribute Lexical Category (LEX):

  • A (adverb)
  • J (adjective)
  • N (noun)
  • V (verb)
  • D (determiner)

etc.
The symbol # is used for words not found in the dictionary. The punctuation is preserved.

Ambiguities

LSS brings all possible categories of a given string. Ambiguities must be represented between {} and isolated by |. They can be local (affecting the whole sentence) or local (affecting only part of it).

  • {A|N} = a string that can be an adverb and a noun
  • {A|J|N} = a string that can be an adverb, an adjective and a noun
  • {N|V} = a string that can be a noun and a verb

etc.

Example

SENTENCE: The book is on the table.
LSS: D {N|V} {V|I} {P|A|N} D {N|V}.

Because:

  • "the" is a determiner
  • "book" may be a verb or a noun
  • "is" may be a verb or an auxiliary
  • "on" may be a preposition, an adverb or a noun
  • "table" may be a noun or a verb

Note that the punctuation (blank spaces and punctuation signs) is preserved in the LSS.

Goal

The main goal of the LSS is to induce disambiguation rules.

Software