# US 2023/0410682 A1 · Task-specific language sets for multilingual learning

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

- Number: US 2023/0410682 A1
- Document ID: US20230410682A1
- Status: Published application
- Kind: A1
- Application number: 17/807,744
- Filed: 2022-06-20
- Published: 2023-12-21
- Earliest claimed: 2022-06-20
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Tejas Indulal Dhamecha; Samarth Bharadwaj; Rudra Murthy Venkataramana; Jaydeep Sen; Anupama Ray
- Web page: https://patentgrep.com/patent/US20230410682A1

## CPC

- G06F 40/40: Processing or translation of natural language (natural language analysis G06F40/20; semantic analysis G06F40/30)
- G09B 19/06: Foreign languages (with audible presentation of material to be studied G09B5/04)
- G06F 40/279: Recognition of textual entities
- G06F 40/30: Semantic analysis
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G09B: EDUCATIONAL OR DEMONSTRATION APPLIANCES; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND, DEAF OR MUTE; MODELS; PLANETARIA; GLOBES; MAPS; DIAGRAMS (devices for psychotechnics or for testing reaction times A61B5/16; games, sports, amusements A63; projectors, projector screens G03B)

## Abstract

A method, a structure, and a computer system for multilingual learning. The exemplary embodiments may include training, for each language in a set of two or more languages, a model for a task and identifying one or more important words appearing in at least two of the models. The exemplary embodiments may further include weighting one or more conflicts and one or more overlaps between the one or more important words, as well as generating a cluster of at least two languages of the set based on an aggregate of the weighting.

## Claims

### Claim 1

A method for multilingual learning, the method comprising:
training, for each language in a set of two or more languages, a model for a task;
identifying one or more important words appearing in at least two of the models;
weighting one or more conflicts and one or more overlaps between the one or more important words; and
generating a cluster of at least two languages of the set based on an aggregate of the weighting.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
utilizing the cluster in one or more higher-order tasks selected from the group consisting of text classification, natural language inference, paraphrase detection, and Multilingual LM.

### Claim 3 (depends on 1)

The method of claim 1, wherein the two or more languages are labelled for the task.

### Claim 4 (depends on 1)

The method of claim 1, wherein the training is via fine-tuning a pretrained model.

### Claim 5 (depends on 1)

The method of claim 1, wherein the task is selected from the group consisting of sentiment classification, natural language interference, and part-of-speech tagging.

### Claim 6 (depends on 1)

The method of claim 1, wherein the weighting the one or more conflicts and the one or more overlaps is based on the task.

### Claim 7 (depends on 1)

The method of claim 1, wherein the identifying one or more important words appearing in at least two of the models is based on the task.

### Claim 8

A computer program product for multilingual learning, the computer program product comprising:
one or more non-transitory computer-readable storage media and program instructions stored on the one or more non-transitory computer-readable storage media capable of performing a method, the method comprising:
training, for each language in a set of two or more languages, a model for a task;
identifying one or more important words appearing in at least two of the models;
weighting one or more conflicts and one or more overlaps between the one or more important words; and
generating a cluster of at least two languages of the set based on an aggregate of the weighting.

### Claim 9 (depends on 8)

The computer program product of claim 8, further comprising:
utilizing the cluster in one or more higher-order tasks selected from the group consisting of text classification, natural language inference, paraphrase detection, and Multilingual LM.

### Claim 10 (depends on 8)

The computer program product of claim 8, wherein the two or more languages are labelled for the task.

### Claim 11 (depends on 8)

The computer program product of claim 8, wherein the training is via fine-tuning a pretrained model.

### Claim 12 (depends on 8)

The computer program product of claim 8, wherein the task is selected from the group consisting of sentiment classification, natural language interference, and part-of-speech tagging.

### Claim 13 (depends on 8)

The computer program product of claim 8, wherein the weighting the one or more conflicts and the one or more overlaps is based on the task.

### Claim 14 (depends on 8)

The computer program product of claim 8, wherein the identifying one or more important words appearing in at least two of the models is based on the task.

### Claim 15

A computer system for multilingual learning, the system comprising:
one or more computer processors, one or more computer-readable storage media, and program instructions stored on the one or more of the computer-readable storage media for execution by at least one of the one or more processors capable of performing a method, the method comprising:
training, for two or more languages, one or more models for a task;
identifying one or more important words within the one or more models;
weighting one or more conflicts and one or more overlaps between the one or more important words across the two or more languages; and
generating a cluster of at least two of the two or more languages based on an aggregate of the weighting.

### Claim 16 (depends on 15)

The computer system of claim 15, further comprising:
utilizing the cluster in one or more higher-order tasks selected from the group consisting of text classification, natural language inference, paraphrase detection, and Multilingual LM.

### Claim 17 (depends on 15)

The computer system of claim 15, wherein the two or more languages are labelled for the task.

### Claim 18 (depends on 15)

The computer system of claim 15, wherein the training is via fine-tuning a pretrained model.

### Claim 19 (depends on 15)

The computer system of claim 15, wherein the task is selected from the group consisting of sentiment classification, natural language interference, and part-of-speech tagging.

### Claim 20 (depends on 15)

The computer system of claim 15, wherein the identifying one or more important words appearing in at least two of the models and the weighting the one or more conflicts and the one or more overlaps is based on the task.
