Updating a language model based on a function-word to total-word ratio
AssigneeInternational Business Machines Corporation
InventorJames R. Lewis
Life of this patent
Estimating the term…
Abstract
Sample text is evaluated to determine if it is suitable for updating a language file in a computer speech recognition system. The evaluation is performed by retrieving from a memory a sample text file containing text and a function word file containing a list of function words. The sample text file is processed to determine a total number of words it contains and a total number of function words it contains. A ratio is calculated for the total number of function words relative to the total number of words. Based upon the value of the ratio thus calculated a decision can then be made as to whether to update a language model.
Claims11
In a computer system for speech recognition, a method for evaluating sample text comprising the steps of:
- retrieving from a memory a sample text file containing text;
- retrieving from said memory a function word file containing a list of function words;
- processing said sample text file to determine a total number of words and a total number of function words contained therein;
- calculating a ratio of said total number of function words to said total number of words;
- and
- updating a language model.
A system for automatically evaluating a sample text to determine if it is suitable for updating a language model in a computer speech recognition system comprising:
- means for retrieving from a computer memory a sample text file containing text;
- a function word file containing a list of function words;
- processing means for analyzing said sample text file to determine a total number of words and a total number of function words contained therein;
- calculating means for calculating a ratio of said total number of function words to said total number of words;
- and
- updating means for updating a language model responsive to said calculating means.
A machine readable storage, having stored thereon a computer program having a plurality of code sections executable by a machine for causing the machine to perform the steps of:
- retrieving from a memory a sample text file containing text;
- retrieving from said memory a function word file containing a list of function words;
- processing said sample text file to determine a total number of words and a total number of function words contained therein;
- calculating a ratio of said total number of function words to said total number of words;
- and
- updating a language model.
Dependencies are inferred from the claim text.