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

Hierarchichal language models

A language model and model technology, applied in speech analysis, speech recognition, instruments, etc., can solve problems such as insufficient data language model and inability to accurately process speech.

Inactive Publication Date: 2004-10-06
NUANCE COMM INC
View PDF0 Cites 15 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Although these systems may contain one or more language models to process user responses, insufficient data may be used to build language models tailored to specific prompts
The result is that this language model becomes too specialized to accurately process the received speech
Specifically, such language models lack the ability to distill from language models to handle more general user responses

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
  • Hierarchichal language models
  • Hierarchichal language models
  • Hierarchichal language models

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0021] The invention disclosed herein relates to a method of constructing a hierarchy of context models and utilizing these context models to convert speech to text. The method of this embodiment can be used in speech recognition systems and dialogue-based natural language understanding systems. Specifically, the present invention can construct multiple context models from different user language sessions, documents, partial documents, and responses in the form of user utterances. These context models can be organized or grouped into related pairs in a bottom-up fashion using known distance scales. Notably, composing context models into related pairs can be achieved automatically and dynamically at runtime. Related pairs of context models can be combined to form a parent context model. This process can be repeated until a hierarchical tree-like structure of the context model is formed. The hierarchy may have a single root node from which other nodes extend. Note that each ...

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 invention disclosed herein concerns a method of converting speech to text using a hierarchy of contextual models. The hierarchy of contextual models can be statistically smoothed into a language model. The method can include processing text with a plurality of contextual models. Each one of the plurality of contextual models can correspond to a node in a hierarchy of the plurality of contextual models. Also included can be identifying at least one of the contextual models relating to the text and processing subsequent user spoken utterances with the identified at least one contextual model.

Description

technical field [0001] The present invention relates to the field of speech recognition and dialog-based systems, and more particularly to the conversion of speech to text using language models. Background technique [0002] Speech recognition is the process of using a computer to convert an acoustic signal received by a microphone into a set of text, numbers or symbols. These recognized words can then be used in various computer software applications for, for example, document preparation, data entry, and prompting and control. The development of speech recognition technology provides an important way to improve user efficiency. [0003] Speech recognition systems can model and classify acoustic signals to form sound models, which are representations of basic language units called phonemes. Upon receiving the acoustic signals, the speech recognition system analyzes the speech signals, recognizes a series of speech models within the acoustic signals, and derives a list of ...

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): G06F17/28G10L15/18
CPCG10L15/183G10L15/197G10L15/26
Inventor 马克·E·爱普斯坦
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