Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Speech recognition by statistical language using square-rootdiscounting

A language and training corpus technology, applied in speech recognition, speech analysis, instruments, etc., can solve the problems of underestimating the probability of events with large counts, prolonging calculation time, rounding errors, etc., and achieve the effect of reliable statistical language modeling

Inactive Publication Date: 2008-02-13
NUANCE COMM INC
View PDF0 Cites 8 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Also, linear discounting shows a strong tendency to underestimate the probability of events seen with large counts, while absolute discounting does not precisely judge the probability of events seen with small counts
[0013] It should be pointed out that in traditional methods, very small numbers often have to be dealt with, which may not only lead to imprecise calculations, such as rounding errors, but may also prolong the calculation time

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Speech recognition by statistical language using square-rootdiscounting
  • Speech recognition by statistical language using square-rootdiscounting
  • Speech recognition by statistical language using square-rootdiscounting

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0073] According to the example shown in Fig. 1, the speaker utters a sentence including three consecutive words a, b and c. The utterance is detected by the microphone 1 and the corresponding microphone signal has been digitized and input into the speech recognition device. The speech recognition device accesses a database comprising training corpus, such as trigrams and / or bigrams found in a large number of novels or radio news broadcast scripts.

[0074] Assume that the speech recognition device has recognized two initial words a and b2 uttered by the speaker. Then the task is to predict the next word of the considered trigram. Based on the training corpus, N possible trigrams (events) e starting from words a and b are known 1 to eN . Every trigram e j (j=1,..,N) is found in the corpus with a frequency of (number of counts) c j .

[0075] In order to predict word c to complete the trigram, the speech recognition device calculates the probability of different candidate...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The present invention relates to a method for statistical language modeling and speech recognition providing a predetermined number of words in a predetermined order, providing a training corpus comprising a predetermined number of sequences of words, wherein each of the sequences consist of the provided predetermined number of words in the predetermined order followed by at least one additional word, providing word candidates and for each word candidate calculating on the basis of the training corpus the probability that the word candidate follows the provided predetermined number of words, determining at least one word candidate for which the calculated probability exceeds a predetermined threshold, wherein the probabilities of the word candidate are calculated based on smoothed maximum-likelihood probabilities calculated for the sequences of words of the training corpus, wherein the smoothed maximum-likelihood probabilities are larger than or equal to a predetermined real positive number that is less than or equal to the inverse of the predetermined number of sequences of words of the training corpus.

Description

technical field [0001] The present invention relates to methods of statistical language modeling using statistical smoothing. In particular, the present invention relates to speech recognition methods based on statistical language modeling using smoothed probability calculations, in particular discounting the probabilities of observed events. Background technique [0002] Statistical language modeling is an attempt to capture the regularities of natural language, and thus the fundamental components of natural language systems for human-computer interaction. Statistical language models aim to estimate the distribution of natural language as precisely as possible. These models play an important role in different natural language applications such as speech recognition, machine translation, text-to-speech systems, and spelling correction. [0003] Speech recognition can be regarded as a particularly important application of statistical language modeling. Speech recognition s...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G10L15/06G10L15/18G10L15/183G10L15/197
CPCG10L15/183G10L15/197G10L15/14G10L15/06
Inventor G·维尔申
Owner NUANCE COMM INC
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com