Instance Model Creation from Unstructured Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing meta-modeling environments lack a unified approach for creating instance models for various domains such as infrastructure, forms, invoices, and clinical trials, and struggle with unstructured data, entity mapping, and attribute mapping, which hinders analysis and automation in fields like healthcare and compliance.
Innovation Solution
A system and method that uses a user interface to provide data sources and an existing ER model to generate a template model, extract information, and create instance models, which can then be merged with existing models, employing techniques like natural language processing and image processing for data extraction and validation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a generic unified approach is used for creating instance models, then adaptability and versatility improve, but device complexity increases
Solution Approach 1:
The patent implements a universal instance model creation framework that can handle multiple data sources (logs, HTML files, PDF files, domain corpus, webs, scanned images) and various domains (infrastructure, forms, invoices, purchase orders, goods received notes, clinical trials, processes) through a single integrated system. The framework uses standardized modules for information extraction, template model generation, and instance model creation that can be applied across different domains, achieving multi-functionality without requiring separate specialized systems for each data type or domain.
Solution Approach 2:
The patent introduces template models as intermediary structures that bridge the gap between diverse data sources and domain-specific instance models. The template model generation module creates intermediate template models that serve as mediators, transforming various input formats into a standardized representation that can then be instantiated into domain-specific models. This intermediary layer simplifies the overall process by providing a common intermediate format that handles the complexity of diverse inputs.
2Measurement precision
If information extraction is performed from unstructured data sources, then measurement precision improves, but difficulty of detecting and measuring increases
Solution Approach 1:
The patent divides the information extraction process into distinct modular stages: information extraction module, template model generation module, and instance model generation module. Each module handles specific aspects of the extraction process, breaking down the complex task of extracting information from unstructured data into manageable segments. This segmentation allows each module to specialize in specific extraction challenges, improving overall precision while making the detection process more systematic and less overwhelming.
Solution Approach 2:
The template model serves as an intermediary structure that facilitates precise information extraction from unstructured data. The template model generation module creates domain-specific template models that act as mediators between the raw unstructured data and the final instance models. These template models encode domain knowledge and extraction rules, making the detection and measurement of relevant information more precise by providing a structured framework for interpretation.
3Manufacturing precision
If entity mapping and attribute mapping are performed, then manufacturing precision improves, but device complexity increases
Solution Approach 1:
The patent combines entity mapping and attribute mapping functions into the template model generation and instance model generation processes. Rather than treating mapping as a separate complex step, the system integrates mapping operations into the template model creation workflow. The template models inherently contain entity-attribute relationships, so mapping occurs naturally during template generation and instance creation, reducing overall system complexity while maintaining high mapping precision.
Solution Approach 2:
The system performs preliminary entity mapping and attribute mapping during the template model generation phase, before actual instance model creation. By pre-establishing the mapping relationships in the template models, the system avoids having to perform complex mapping operations repeatedly during instance generation. This preliminary action captures the mapping logic once in the template layer, simplifying subsequent instance creation while ensuring consistent high-precision mapping across all instances.
Data Source
AI summary
A system and method for creating an instance model is provided. The system provides an information extraction and modeling framework from wide spectrum of document types such as PDF, Text, HTML, LOG, CSV, images, audio/video files and DOCX. In this framework information is extracted and mapped on a domain conceptual model like ER model and the instance model is created. Initially a template model is created using the existing ER model and the plurality of data sources. The template model, the existing ER model and the information extracted from the plurality of data sources are then provided as input to generate the instance model. The system or method is not limited to extract information from log files. This can be useful for different types of files type if the structures and formats of data are different. The system can also be used with unstructured type of data sources.