Business Process Structuring via Entity Grouping and Procedure Ordering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for generating structured data from documents describing business processes require large amounts of learning data and precise rules, making them difficult to apply in fields with limited data.
Innovation Solution
A computer system that extracts entities from documents, classifies their categories, groups them into procedures, identifies main entities, determines the order of procedures based on relationships, and generates structured data without relying on precise rules.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If machine learning techniques are used to generate structured data from documents, then automation is improved, but large amounts of learning data are required
Solution Approach 1:
The patent segments the document structuring task into multiple independent components: entity extraction, relationship identification, and structure generation. Each component can be processed separately using simple rules or templates, avoiding the need for large-scale machine learning models that require extensive training data.
Solution Approach 2:
The patent introduces an intermediary structured format (such as JSON or XML templates) that bridges the gap between unstructured document text and the desired structured output. This intermediary structure guides the extraction and organization process, reducing dependency on complex learning algorithms.
2Quantity of substance
If rule-based techniques are used to generate structured data, then learning data requirement is reduced, but precise rules must be set
Solution Approach 1:
The patent creates universal templates and patterns that can be applied across different document types and domains. These templates define common structural patterns (e.g., process steps, relationships between entities) that work universally, reducing the need to create domain-specific rules for each case.
Solution Approach 2:
The patent employs dynamic rule selection where the applicable rules and templates are determined based on the characteristics of the input document. The system adapts which rules to apply by analyzing document type, content, and structure, making the rule-based system more flexible and easier to maintain.
3Measurement precision
If manual methods are used to generate structured data, then accuracy is improved, but time consumption increases
Solution Approach 1:
The patent enables the system to automatically perform tasks that would otherwise require manual intervention. By using entity extraction algorithms and relationship identification rules, the system self-services the structured data generation process, achieving both speed and acceptable accuracy without human involvement.
Solution Approach 2:
The patent incorporates feedback mechanisms where the generated structured data can be validated against the original document and common sense constraints. This feedback loop ensures accuracy by detecting and correcting errors in the automated generation process.
Data Source
AI summary
Provided is a computer system that extracts an entity from a document that describes a business process including a plurality of procedures, and that classifies a category of the entity. The computer system generates a plurality of entity groups each including one or more entities and corresponding to one procedure, and specifies for each of the entity groups, a main entity that is the entity, which characterizes a procedure corresponding to the entity group, based on a category of one or more of the entities included in the entity group. The computer system executes processing of determining an order of the plurality of procedures based on a relationship between main entities, determines an order of the plurality of procedures based on a result of the processing, and generates information related to the ordered entity groups as structured data of the business process.