# US 12,602,404 B2 · Data labeling work support apparatus, data labeling work support 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 12,602,404 B2
- Document ID: US12602404B2
- Status: Granted
- Kind: B2
- Application number: 18/175,762
- Filed: 2023-02-28
- Published: 2024-03-14
- Granted: 2026-04-14
- Earliest claimed: 2022-09-14
- Assignees: Kabushiki Kaisha Toshiba
- Applicants: Kabushiki Kaisha Toshiba
- Inventors: Kouta Nakata; Kentaro Takagi; Yaling Tao
- Web page: https://patentgrep.com/patent/US12602404B2

## CPC

- G06F 16/287: {Visualization; Browsing}
- G06F 16/54: Browsing; Visualisation therefor
- G06F 16/55: Clustering; Classification
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

According to one embodiment, a data labeling work support apparatus includes a processor including hardware. The processor acquires a first label assigned to data. The processor acquires the data. The processor extracts a feature of the data. The processor groups the data based on a similarity or a distance of the feature. The processor assigns a second label to the grouped data. The processor calculates a degree of matching between the first label and the second label. The processor outputs information regarding a combination of the first label and the second label having a low degree of matching.

## Claims

### Claim 1

A data labeling work support apparatus comprising:
a processor including hardware; and
a memory configured to store instructions that, when executed by the processor, cause the processor to:
acquire training data including a plurality of data pieces;
acquire a plurality of first labels manually assigned to the data pieces of the training data,
extract features of the data pieces of the training data by using a first unsupervised learning method without labels,
group the data pieces of the training data based on a similarity or a distance of the features by using a second unsupervised learning method,
assign a plurality of second labels to the grouped data pieces of the training data by using the second unsupervised learning method,
calculate, for each of the data pieces, a degree of matching between the first label assigned to the data piece and the second label assigned to the data piece,
display, on a screen of a display device, a plurality of plot points representing the training data using the features, each plot point corresponding to one data piece, and
highlight a plot point representing a data piece of the training data corresponding to a combination of the first label and the second label having a low degree of matching.

### Claim 2 (depends on 1)

The data labeling work support apparatus according to claim 1, wherein the number of groups to be grouped is set based on the number of types of the first labels.

### Claim 3 (depends on 1)

The data labeling work support apparatus according to claim 1, wherein
the processor receives a signal corresponding to an operation for selecting a plot point displayed on the screen, and assigns a plurality of third labels to a first data piece corresponding to the selected plot point and a plurality of second data pieces respectively corresponding to a plurality of plot points in a vicinity of the selected plot point,
the processor calculates, for each of the first data piece and the second data pieces, a degree of matching between the first label assigned to the data piece and the third label assigned to the data piece, and
the processor highlights a plot point representing a data piece among the first data piece and the second data pieces corresponding to a combination of the first label and the third label having a low degree of matching.

### Claim 4 (depends on 1)

The data labeling work support apparatus according to claim 1, wherein
the processor receives a signal corresponding to an operation for selecting the highlighted plot point displayed on the screen, the selected plot point corresponding to a first data piece,
the processor lists a second data piece having a same combination of the first label and the second label as a combination of the first label and the second label assigned to the first data piece corresponding to the selected plot point, and
the processor highlights a plot point corresponding to the listed second data piece.

### Claim 5 (depends on 1)

The data labeling work support apparatus according to claim 1, wherein
the processor acquires, for each of the data pieces, a plurality of first labels manually assigned by a plurality of persons,
the processor generates, for each of the data pieces, a combined label in which the first labels assigned by the plurality of persons are combined,
the processor calculates, for each of the data pieces, a degree of matching between the combined label and the second label, and
the processor highlights a plot point representing a data piece corresponding to a combination of the combined label and the second label having a low degree of matching.

### Claim 6

A data labeling work support method comprising:
acquiring training data including a plurality of data pieces;
acquiring a plurality of first labels manually assigned to the data pieces of the training data;
extracting features of the data pieces of the training data by using a first unsupervised learning method without labels;
grouping the data pieces of the training data based on a similarity or a distance of the features by using a second unsupervised learning method;
assigning a plurality of second labels to the grouped data pieces of the training data by using the second unsupervised learning method;
calculating, for each of the data pieces, a degree of matching between the first label assigned to the data piece and the second label assigned to the data piece;
displaying, on a screen of a display device, a plurality of plot points representing the training data using the features, each plot point corresponding to one data piece; and
highlighting a plot point representing a data piece of the training data corresponding to a combination of the first label and the second label having a low degree of matching.

### Claim 7

A non-transitory computer readable storage medium for storing a data labeling work support program for causing a computer to execute processing comprising:
acquiring training data including a plurality of data pieces;
acquiring a plurality of first labels manually assigned to the data pieces of the training data;
extracting a features of the data pieces of the training data by using a first unsupervised learning method without labels;
grouping the data pieces of the training data based on a similarity or a distance of the features by using a second unsupervised learning method;
assigning a plurality of second labels to the grouped data pieces of the training data by using the second unsupervised learning method;
calculating, for each of the data pieces, a degree of matching between the first label assigned to the data piece and the second label assigned to the data piece;
displaying, on a screen of a display device, a plurality of plot points representing the training data using the features, each plot point corresponding to one data piece; and
highlighting a plot point representing a data piece of the training data corresponding to a combination of the first label and the second label having a low degree of matching.