METHOD AND SYSTEM FOR AUTOMATIC WORD SPACING OF VOICE RECOGNITION USING NAMED ENTITY RECOGNITION

摘要

The present invention provides a method and a system for compensating voice recognized word spacing using the recognition of an entity name, wherein the method comprises: a step of recognizing an inputted voice and generating voice text; an error section determination step of estimating an error section of the voice recognition via a natural language processing procedure for the voice text, and setting the error section as a compensation target; a category estimation step of extracting the compensation target and a usage pattern using the compensation target in the context of the compensation target from the voice text, and comparing the extracted usage pattern with the usage pattern of an object stored in an entity name usage pattern database for each category to estimate the category corresponding to the compensation target; and a spacing compensation step of analyzing an occurrence frequency of a syllable N-gram for each category for the compensation target based on an entity name dictionary database for each category to compensate the spacing of the compensation target. The present invention first checks a scope of a corresponding word with the category even though words not registered in a voice recognition dictionary such as a proper noun, a new-coined word, a varied word or the like and applying the checked scope to a spacing probability, thereby compensating the spacing error of the voice recognition, correctly and reliably.