Grammar fragment acquisition using syntactic and semantic clustering
AssigneeAT&T Corp.
InventorsKazuhiro Arai, Allen Louis Gorin, Giuseppe Riccardi, Jeremy Huntley Wright
Life of this patent
Estimating the term…
Abstract
A method and apparatus are provided for automatically acquiring grammar fragments for recognizing and understanding fluently spoken language. Grammar fragments representing a set of syntactically and semantically similar phrases may be generated using three probability distributions: of succeeding words, of preceding words, and of associated call-types. The similarity between phrases may be measured by applying Kullback-Leibler distance to these three probability distributions. Phrases being close in all three distances may be clustered into a grammar fragment.
Claims23
A method for acquiring grammar fragments, comprising:
- selecting candidate phrases from a set of words, phrases or symbols;
- measuring semantic and syntactic similarity in the candidate phrases;
- and
- clustering the candidate phrases into grammar fragments based on the semantic and syntactic similarity measurements,
- wherein the method is recursive.
An apparatus that acquires grammar fragments, comprising:
- a candidate phrase selector that selects candidate phrases from a set of words, phrases or symbols;
- a distance calculation device that measures the semantic and syntactic similarity in the candidate phrases selected by the candidate phrase selector;
- and
- a grammar fragment clustering device that clusters the selected candidate phrases into grammar fragments based on the semantic and syntactic similarity measurements made by the distance calculation device.
An apparatus that automatically acquires grammar fragments, comprising:
- candidate phrase selecting means for selecting candidate phrases from a set of words, phrases or symbols;
- distance calculation means for measuring the semantic and syntactic similarity in the candidate phrases selected by the candidate phrase selecting means;
- and
- grammar fragment clustering means for clustering the selected candidate phrases into grammar fragments based on the semantic and syntactic similarity measurements made by the distance calculation means.
Dependencies are inferred from the claim text.