# US 2024/0004907 A1 · Knowledge graph question answering with neural machine translation

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

- Number: US 2024/0004907 A1
- Document ID: US20240004907A1
- Status: Published application
- Kind: A1
- Application number: 17/810,013
- Filed: 2022-06-30
- Published: 2024-01-04
- Earliest claimed: 2022-06-30
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Saswati Dana; Dinesh Garg; Dinesh Khandelwal; G P Shrivatsa Bhargav; Sukannya Purkayastha
- Web page: https://patentgrep.com/patent/US20240004907A1

## CPC

- G06F 16/3329: Natural language query formulation
- G06F 40/20: Natural language analysis (semantic analysis of natural language G06F40/30)
- G06F 40/284: Lexical analysis, e.g. tokenisation or collocates
- G06F 40/30: Semantic analysis
- G06F 40/35: Discourse or dialogue representation
- G06N 3/04: Architecture, e.g. interconnection topology
- G06N 3/045: Combinations of networks
- G06N 3/08: Learning methods
- G06N 5/02: Knowledge representation; Symbolic representation
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A modular two-stage neural architecture is used in translating a natural language question into a logic form such as a SPARQL Protocol and RDF Query Language (SPARQL) query. In a first stage, a neural machine translation (NMT)-based sequence-to-sequence (Seq2Seq) model translates a question into a sketch of the desired SPARQL query called a SPARQL silhouette. In a second stage a neural graph search module predicts the correct relations in the underlying knowledge graph.

## Claims

### Claim 1

A computer-implemented method comprising:
receiving an input question for a question-answering (QA) system;
translating the input question to a SPARQL Protocol and RDF Query Language (SPARQL) query using neural machine translation (NMT) and a neural graph search module.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, further comprising using a noise simulator to:
simulate varying levels of noise in an entity/relation linking process; and
to mask mentions and entities/relations in the input question text and the SPARQL query.

### Claim 3 (depends on 2)

The computer-implemented method of claim 2, wherein the noise simulator is configured to simulate a realistic amount of noise by employing a corresponding masking strategy for entities and relations.

### Claim 4 (depends on 2)

The computer-implemented method of claim 2, wherein the noise simulator is configured to simulate an idealistic amount of noise by employing a corresponding masking strategy for entities and relations.

### Claim 5 (depends on 2)

The computer-implemented method of claim 2, wherein the noise simulator is configured to simulate a significant amount of noise by employing a corresponding masking strategy for entities and relations.

### Claim 6 (depends on 1)

The computer-implemented method of claim 1, wherein the neural network graph search module is a Bidirectional Encoder Representations from Transformers (BERT) based graph search module.

### Claim 7 (depends on 1)

The computer-implemented method of claim 1, wherein a sequence-to-sequence (Seq2Seq) model is used to translate the input question to the SPARQL query.

### Claim 8

A system comprising:
a processor; and
a memory in communication with the processor, the memory containing instructions that, when executed by the processor, cause the processor to:
receive an input question for a question-answering (QA) system; and
translate the input question to a SPARQL Protocol and RDF Query Language (SPARQL) query using neural machine translation (NMT) and a neural graph search module.

### Claim 9 (depends on 8)

The system of claim 8, the memory containing additional instructions that, when executed by the processor, cause the processor to use a noise simulator to:
simulate varying levels of noise in an entity/relation linking process; and
mask mentions and entities/relations in the input question text and the SPARQL query.

### Claim 10 (depends on 9)

The system of claim 9, wherein the noise simulator is configured to simulate a realistic amount of noise by employing a corresponding masking strategy for entities and relations.

### Claim 11 (depends on 9)

The system of claim 9, wherein the noise simulator is configured to simulate an idealistic amount of noise by employing a corresponding masking strategy for entities and relations.

### Claim 12 (depends on 9)

The system of claim 9, wherein the noise simulator is configured to simulate a significant amount of noise by employing a corresponding masking strategy for entities and relations.

### Claim 13 (depends on 8)

The system of claim 8, wherein the neural network graph search module is a Bidirectional Encoder Representations from Transformers (BERT) based graph search module.

### Claim 14 (depends on 8)

The system of claim 8, wherein a sequence-to-sequence (Seq2Seq) model is used to translate the input question to the SPARQL query.

### Claim 15

A computer program product, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to:
receive an input question for a question-answering (QA) system; and
translate the input question to a SPARQL Protocol and RDF Query Language (SPARQL) query using neural machine translation (NMT) and a neural graph search module.

### Claim 16 (depends on 15)

The computer program product of claim 15, the computer readable storage medium having additional program instructions embodied therewith, the program instructions executable by the computer to further cause the computer to use a noise simulator to:
simulate varying levels of noise in an entity/relation linking process; and
mask mentions and entities/relations in the input question text and the SPARQL query.

### Claim 17 (depends on 16)

The computer program product of claim 16, wherein the noise simulator is configured to simulate an idealistic amount of noise by employing a corresponding masking strategy for entities and relations.

### Claim 18 (depends on 16)

The computer program product of claim 16, wherein the noise simulator is configured to simulate a realistic amount of noise by employing a corresponding masking strategy for entities and relations.

### Claim 19 (depends on 16)

The computer program product of claim 16, wherein the noise simulator is configured to simulate a significant amount of noise by employing a corresponding masking strategy for entities and relations.

### Claim 20 (depends on 15)

The computer program product of claim 15, wherein the neural network graph search module is a Bidirectional Encoder Representations from Transformers (BERT) based graph search module.
