# US 2019/0378496 A1 · Recognition device, method and storage medium

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

- Number: US 2019/0378496 A1
- Document ID: US20190378496A1
- Status: Published application
- Kind: A1
- Application number: 16/429,911
- Filed: 2019-06-03
- Published: 2019-12-12
- Earliest claimed: 2018-06-07
- Assignees: Kabushiki Kaisha Toshiba; Toshiba Digital Solutions Corporation
- Applicants: Kabushiki Kaisha Toshiba; Toshiba Digital Solutions Corporation
- Inventors: Koji YASUDA; Kenta Cho
- Web page: https://patentgrep.com/patent/US20190378496A1

## CPC

- G10L 15/06: Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice (G10L15/14 takes precedence)
- G10L 15/083: {Recognition networks (G10L15/142, G10L15/16 take precedence)}
- G06F 40/205: Parsing
- G06F 40/284: Lexical analysis, e.g. tokenisation or collocates
- G10L 15/26: Speech to text systems (G10L15/08 takes precedence)
- G10L 15/32: Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems
- G10L 2015/088: {Word spotting}
- 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

According to one embodiment, a recognition device includes storage and a processor. The storage is configured to store a first recognition model, a first data set, and tags, for each first recognition model. The processor is configured to acquire a second data set, execute recognition processing of the second recognition target data in the second data set by using the first recognition model, extract a significant tag of the tags stored in the storage in association with the first recognition model, based on the recognition processing result and the second correct data in the second data set, and create a second recognition model based on the acquired second data set and the first data set stored in the storage in association with the extracted tag.

## Claims

### Claim 1

A recognition device, comprising:
storage configured to store a first recognition model created based on a first data set comprising first recognition target data collected under a predetermined condition and first correct data to be recognized from the first recognition target data, the first data set, and tags indicative of the condition, for each first recognition model; and
a processor configured to:
acquire a second data set comprising second recognition target data and second correct data to be recognized from the second recognition target data;
execute recognition processing of the second recognition target data in the second data set by using the first recognition model stored in the storage;
extract a significant tag of the tags stored in the storage in association with the first recognition model used for the recognition processing, based on the recognition processing result and the second correct data in the second data set; and
create a second recognition model based on the acquired second data set and the first data set stored in the storage in association with the extracted tag.

### Claim 2 (depends on 1)

The recognition device of claim 1, wherein
the first recognition target data in the first data set comprises first voice data,
the first correct data in the first data set comprises a first text written from the first voice data,
the second recognition target data in the second data set comprises second voice data,
the second correct data in the second data set comprises a second text written from the second voice data, and
the recognition processing comprises processing of recognizing voice from the voice data and converting the voice into a text.

### Claim 3 (depends on 2)

The recognition device of claim 2, wherein the processor is configured to:
input third voice data, and
convert the third voice data into a third text with the created second recognition model.

### Claim 4 (depends on 2)

The recognition device of claim 2, wherein the processor is configured to:
compute a recognition accuracy of the first recognition model by comparing a recognition processing result of the second voice data using the first recognition model with the second text;
compute a statistic with the computed recognition accuracy of the first recognition model; and
extract a tag of significantly high recognition accuracy from the tags stored in the storage in association with the first recognition model.

### Claim 5 (depends on 2)

The recognition device of claim 2, wherein the processor is configured to:
display the extracted tag; and
create a second recognition model based on the acquired second dataset and the first data set stored in the storage in association with a tag designated by the user, of the displayed tags.

### Claim 6 (depends on 5)

The recognition device of claim 5, wherein the processor is configured to display a tag related to the extracted tag.

### Claim 7 (depends on 2)

The recognition device of claim 2, wherein the processor is configured to:
execute recognition processing of the second voice data in the second data set by using the created second recognition model;
compute a recognition accuracy of the second recognition model by comparing the recognition processing result of the second voice data using the created second recognition model with the second text; and
display the computed recognition accuracy of the second recognition model.

### Claim 8 (depends on 7)

The recognition device of claim 7, wherein the processor is configured to:
create a third recognition model based on the second data set;
execute recognition processing of the second voice data in the second data set by using the created third recognition model;
compute a recognition accuracy of the third recognition model by comparing the recognition processing result of the second voice data using the created third recognition model with the second text in the second data set; and
display the recognition accuracy of the second recognition model and the recognition accuracy of the third recognition model.

### Claim 9 (depends on 2)

The recognition device of claim 2, wherein the processor is configured to:
determine whether a data amount of the acquired second data set and the first data set stored in the storage in association with the extracted tag is sufficient or not;
create a third data set based on the second data set if it is determined that the data amount of the second data set and the first data set is not sufficient;
create a second recognition model, based on the acquired second data set, the first data set stored in the storage in association with the extracted tag, and the created third data set.

### Claim 10 (depends on 9)

The recognition device of claim 9, wherein the processor is configured to:
create third voice data from a third text acquired based on a keyword extracted from the second text in the second data set; and
create a third data set comprising the third voice data and the third text.

### Claim 11

A method executed by a recognition device comprising storage configured to store a first recognition model created based on a first data set comprising first recognition target data collected under a predetermined condition and first correct data to be recognized from the first recognition target data, the first data set, and tags indicative of the condition, for each first recognition model, the method comprising;
acquiring a second data set comprising second recognition target data and second correct data to be recognized from the second recognition target data;
executing recognition processing of the second recognition target data in the second data set by using the first recognition model stored in the storage;
extracting a significant tag of the tags stored in the storage in association with the first recognition model used for the recognition processing, based on the recognition processing result and the second correct data in the second data set; and
creating a second recognition model based on the acquired second data set and the first data set stored in the storage in association with the extracted tag.

### Claim 12

A non-transitory computer-readable storage medium having stored thereon a computer program which is executable by a computer using storage configured to store a first recognition model created based on a first data set comprising first recognition target data collected under a predetermined condition and first correct data to be recognized from the first recognition target data, the first data set, and tags indicative of the condition for first recognition model, the computer program comprising instructions capable of causing the computer to execute functions of:
acquiring a second data set comprising second recognition target data and second correct data to be recognized from the second recognition target data;
executing recognition processing of the second recognition target data in the second data set by using the first recognition model stored in the storage;
extracting a significant tag of the tags stored in the storage in association with the first recognition model used for the recognition processing, based on the recognition processing result and the second correct data in the second data set; and
creating a second recognition model based on the acquired second data set and the first data set stored in the storage in association with the extracted tag.
