Browse Prior Art Database

Sense Checker

IP.com Disclosure Number: IPCOM000013368D
Original Publication Date: 2001-Apr-13
Included in the Prior Art Database: 2003-Jun-18

Publishing Venue

IBM

Abstract

This proposal concerns systems which process text in spoken or written form, such as word processors or voice recognition software. Typically, the text typed into word processors or spoken into voice recognition software is checked for correctness with a spell checker. In addition a grammar checker might check the text for grammatical correctness. However, both types of checkers usually miss a lot of typographical and/or semantical errors because they don't understand the meaning of the text to be checked. The proposed sense checker is a system that understands the meaning of the text typed into a word processor or spoken into a voice recognition software. It can be based on MeanX or any other meaning recognition tool as a system to understand meaning in natural language. As a result of this understanding process, words of the input get classified into a knowledge repository according to their assumed meaning from the context and according to their spelling. If a word has been misspelled (and even if the misspelled word exists) or if a word has been used in an incorrect sense, this will be noticed by the system because of a bad match during the classification process. In particular, there will be a discrepancy between the classification neighborhood in the knowledge repository for the word with respect to its context and with respect to its spelling. As a result, the system can automatically correct the misspelled word with the correct word, or replace the word that has been used in an incorrect sense with a different word that has the desired meaning.