Entity Identification in Semi-Structured Content Using Ensemble ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Entity identifiers struggle to accurately identify entities in semi-structured content, such as email signature blocks, due to variations in formatting, which hinders the extraction of relevant information like names, addresses, and contact details.
Innovation Solution
An ensemble machine learning model is employed to assign entity types and boundaries to tokens in semi-structured content, utilizing multiple layers of classification, including latent dynamic conditional random fields, latent insight models, and shallow parsing, to accurately identify and extract entities like names, addresses, and contact information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If an entity identifier is used to identify entities in unstructured content, then it can properly identify entities in framed sentences, but it fails to accurately identify entities in semi-structured content like email signature blocks
Solution Approach 1:
The patent segments the entity identification process into multiple specialized components: an unstructured content entity identifier for framed sentences, a semi-structured content entity identifier for email signatures, and a structured content entity identifier for forms. Each component is optimized for specific content types, resolving the contradiction by dividing the monolithic identification task into adaptable specialized segments.
Solution Approach 2:
The system dynamically selects which entity identifier to use based on the content type being processed. The framework adapts its identification approach by switching between unstructured, semi-structured, and structured content identifiers according to the input, enabling high accuracy across varying content formats without requiring a single rigid identifier.
2Ease of manufacture
If a single entity identifier is used for all content types, then the system is simple to implement, but it cannot handle variations in formatting across different semi-structured content
Solution Approach 1:
The patent creates a universal entity identification framework that can handle multiple content types through a single integrated system. The framework provides a common interface and processing pipeline that works across unstructured, semi-structured, and structured content, achieving both implementation simplicity and extraction reliability through multi-functional design.
Solution Approach 2:
The patent introduces intermediary components including a content type detector that mediates between the input and entity identifiers, and a result merger that mediates between multiple identifier outputs. These intermediaries simplify implementation by providing standardization layers while enabling the system to reliably handle diverse content formats.
3Loss of information
If traditional entity identifiers are used, then they work for standard sentences, but they fail to extract relevant information like names, addresses, and contact details from email signature blocks
Solution Approach 1:
The patent applies local quality by creating specialized entity identifiers with specific functionalities for different content types. The semi-structured content entity identifier is specifically optimized to extract names, addresses, and contact details from email signatures, while the unstructured content identifier handles framed sentences. Each identifier has tailored capabilities matching its target content type, ensuring complete information extraction without requiring overly complex universal logic.
Data Source
AI summary
Identifying entities in semi-structured content is described. A system assigns a corresponding entity type based on a corresponding entity type score for each token in a sequence of tokens in semi-structured content, based on multiple entity types, wherein each token is a corresponding character set. The system assigns a corresponding boundary type based on a corresponding boundary type score for each token in the sequence of tokens, based on a begin boundary type or a continue boundary type. The system identifies an entity based on a corresponding entity type score and a corresponding boundary type for each token in the sequence of tokens. The system outputs the sequence of tokens as an identified set of entities based on the identified entity.