Document-based requirements extraction

Patent metadata

Field	Value
Publication number	US-11704489-B2
Application number	US-202016917138-A
Country	US
Kind code	B2
Filing date	Jun 30, 2020
Priority date	Jun 30, 2020
Publication date	Jul 18, 2023
Grant date	Jul 18, 2023

Abstract

A computer-implemented method, system, and computer program product for identifying requirements in a document. A document including requirements is received. Attribute related information in the document is identified using an attribute model. Component information in the document is identified using a component model. The attribute related information and the component information identified in the document are merged. Requirements in the document are identified from the merged attribute related information and component information. The requirements identified in the document are used to develop a product.

First claim

What is claimed is:

1. A computer-implemented method of identifying requirements in a document, comprising:
receiving the document;
identifying component information in the document using a component model;
identifying attribute related information in the document using an attribute model, wherein the attribute model is a machine learning model that is trained using training documents;
merging the component information and the attribute related information identified in the document;
identifying the requirements in the document from the merged component information and attribute related information; and
using the requirements identified in the document to produce a product, wherein the component model is a domain-specific model that is applicable for a specific domain; and the attribute model is a common model that is applicable for all domains; wherein a domain model comprises the component model and the attribute model, and wherein the domain model is continuously improved by user feedback and self-learning.

2. The computer-implemented method of claim 1 further comprising:
receiving user feedback on the requirements identified in the document; and
modifying at least one of the component model and the attribute model in response to the user feedback.

3. The computer-implemented method of claim 1 , wherein the component model comprises:
a general component corpus of general terms for components from a plurality of domains; and
a domain-specific component corpus of domain-specific terms for components for a specific domain.

4. The computer implemented method of claim 3 further comprising generating the component model by:
generating a basic component corpus of terms for components from one or more of documents, websites, and operator input;
distinguishing the terms of the basic component corpus into general terms and specific terms;
calculating inverse document frequency scores of term pairs; and
calculating reusability and finding the part that can be reused for the new domain from the general component corpus based on the inverse document frequency scores.

5. The computer-implemented method of claim 1 , wherein the attribute related information comprises:
an attribute of a component;
a value for the attribute;
a guide word describing a relationship between the value and the attribute; and
a unit of the value.

6. A system for identifying requirements in a document, comprising a data processing system configured to:
receive the document;
identify component information in the document using a component model;
identify attribute related information in the document using an attribute model, wherein the attribute model is a machine learning model that is trained using training documents;
merge the component information and the attribute related information identified in the document;
identify the requirements in the document from the merged component information and attribute related information; and
use the requirements identified in the document to produce a product, wherein the component model is a domain-specific model that is applicable for a specific domain; and the attribute model is a common model that is applicable for all domains; wherein a domain model comprises the component model and the attribute model, and wherein the domain model is continuously improved by user feedback and self-learning.

7. The system of claim 6 , wherein the data processing system is further configured to:
receive user feedback on the requirements identified in the document; and
modify at least one of the component model and the attribute model in response to the user feedback.

8. The system of claim 6 , wherein the component model comprises:
a general component corpus of general terms for components from a plurality of domains; and
a domain-specific component corpus of domain-specific terms for components for a specific domain.

9. The system of claim 8 , wherein the data processing system is configured to generate the component model by:
generating a basic component corpus of terms for components from one or more of documents, websites, and operator input;
distinguishing the terms of the basic component corpus into general terms and specific terms;
calculating inverse document frequency scores of term pairs; and
calculating reusability and finding the part that can be reused for the new domain from the general component corpus based on the inverse document frequency score.

10. The system of claim 6 , wherein the attribute related information comprises:
an attribute of a component;
a value for the attribute;
a guide word describing a relationship between the value and the attribute; and
a unit of the value.

11. A computer program product for identifying requirements in a document, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a device to cause the device to:
receive the document;
identify component information in the document using a component model;
identify attribute related information in the document using an attribute model, wherein the attribute model is a machine learning model that is trained using training documents;
merge the component information and the attribute related information identified in the document;
identify the requirements in the document from the merged component information and attribute related information; and
use the requirements identified in the document to produce a product, wherein the component model is a domain-specific model that is applicable for a specific domain; and the attribute model is a common model that is applicable for all domains; wherein a domain model comprises the component model and the attribute model, and wherein the domain model is continuously improved by user feedback and self-learning.

12. The computer program product of claim 11 , wherein the program instructions are executable by the device to cause the device to:
receive user feedback on the requirements identified in the document; and
modify at least one of the component model and the attribute model in response to the user feedback.

13. The computer program product of claim 11 , wherein the component model comprises:
a general component corpus of general terms for components from a plurality of domains; and
a domain-specific component corpus of domain-specific terms for components for a specific domain.

14. The computer program product of claim 11 , wherein the attribute related information comprises:
an attribute of a component;
a value for the attribute;
a guide word describing a relationship between the value and the attribute; and
a unit of the value.

Assignees

• IBM

Inventors

• Miao Wan Wan
• Zhan sheng hui
• Hu Zhi Lin
• Cao Shu Xia
• Chen jian hui
• Liu Tao

Classifications

• G06F40/279 Primary

  Recognition of textual entities · CPC title

• G06N20/00

  Machine learning · CPC title