# US 2016/0365093 A1 · System and method for automatic language model selection

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2016/0365093 A1
- Document ID: US20160365093A1
- Status: Published application
- Kind: A1
- Application number: 14/736,282
- Filed: 2015-06-11
- Published: 2016-12-15
- Earliest claimed: 2015-06-11
- Assignees: Nice Systems Ltd.
- Applicants: Nice-Systems Ltd.
- Inventors: Maor NISSAN
- Web page: https://patentgrep.com/patent/US20160365093A1

## CPC

- G10L 15/187: Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
- G10L 15/183: using context dependencies, e.g. language models
- G06F 40/242: Dictionaries
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A system and method for generating a transcript of an audio input. An embodiment of a system and method may include generating a phonetic lattice by decoding the audio input and producing a transcription based on the phonetic lattice and based on a first language model. A transcription may be analyzed to produce analysis results. Analysis results may be used to select from a plurality of language models, one language model and the selected language model may be used to generate a transcript of the audio input.

## Claims

### Claim 1

A computer-implemented method of generating a transcript of an audio input, the method comprising:
generating a phonetic lattice by decoding the audio input;
producing a transcription based on the phonetic lattice and based on a first language model;
analyzing the transcription to produce analysis results and selecting, from a plurality of language models and based on the analysis result, a second language model; and
using the second language model to generate a second transcript of the audio input.

### Claim 2 (depends on 1)

The method of claim 1, wherein the first language model is a general model and the second language model is a specific language model.

### Claim 3 (depends on 1)

The method of claim 1, comprising:
associating words identified in the transcription with a certainty value calculated for each identified word;
including words associated with a certainty value higher than a first threshold value in a combined index;
combining at least two sub-words identified in the transcription to produce a word and including the produced word in the combined index; and
using the combined index to select the second language model.

### Claim 4 (depends on 3)

The method of claim 3, comprising identifying key phrases in the transcription and including the key phrases in the combined index.

### Claim 5 (depends on 3)

The method of claim 3, comprising, selecting the second language model based on calculating a plurality of match values by relating a respective plurality of specific language models to the combined index and selecting the specific language model associated with the highest match value.

### Claim 6 (depends on 1)

The method of claim 1, comprising:
associating sub-words identified in the transcription with a certainty value calculated for each identified sub-word;
identifying a sequence of sub-words associated with a certainty value higher than a second threshold value;
producing a word based on the identified sequence of sub-words; and
including the produced word in the combined index.

### Claim 7

A method of generating a transcript of an audio input, the method comprising:
producing a first transcription of the audio input using a first language model;
selecting, from a plurality of language models, a second language model by matching the plurality of language models with the first transcription; and
producing a second transcription of the audio input using the second language model.

### Claim 8

An article comprising a computer-readable storage medium, having stored thereon instructions that, when executed by a controller, cause the controller to:
generate a phonetic lattice by decoding the audio input;
produce a transcription based on the phonetic lattice and based on a first language model;
analyze the transcription to produce analysis results and select, from a plurality of language models and based on the analysis result, a second language model; and
use the second language model to generate a second transcript of the audio input.

### Claim 9 (depends on 8)

The article of claim 8, wherein the first language model is a general model and the second language model is a specific language model.

### Claim 10 (depends on 8)

The article of claim 8, wherein the instructions when executed further result in:
associating words identified in the transcription with a certainty value calculated for each identified word;
including words associated with a certainty value higher than a first threshold value in a combined index;
combining at least two sub-words identified in the transcription to produce a word and including the produced word in the combined index; and
using the combined index to select the second language model.

### Claim 11 (depends on 10)

The article of claim 10, wherein the instructions when executed further result in identifying key phrases in the transcription and including the key phrases in the combined index.

### Claim 12 (depends on 10)

The article of claim 10, wherein the instructions when executed further result in selecting the second language model based on calculating a plurality of match values by relating a respective plurality of specific language models to the combined index and selecting the specific language model associated with the highest match value.

### Claim 13 (depends on 8)

The article of claim 8, wherein the instructions when executed further result in:
associating sub-words identified in the transcription with a certainty value calculated for each identified sub-word;
identifying a sequence of sub-words associated with a certainty value higher than a second threshold value;
producing a word based on the identified sequence of sub-words; and
including the produced word in the combined index.
