Named Entity Extraction Using Structural Markers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing named entity tagging systems are limited in accuracy, language compatibility, and scalability, requiring extensive manual training data and being computationally expensive, making them inefficient for diverse document types and languages.
Innovation Solution
A computer-implemented method identifies marker sequences in input document strings, assigns eligibility values, and generates output data to extract named entities, using proximity strings and eligibility calculations to improve tagging efficiency and language independence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing named entity tagging systems are used, then named entities can be extracted from documents, but the systems require extensive manual training data and are computationally expensive
Solution Approach 1:
The patent extracts and utilizes structural markers (section headings, numbered lists, tables, figures) that are inherently present in technical documents to guide named entity recognition. This eliminates the need for extensive manual training data by leveraging the document's own structure as the training signal, directly resolving the contradiction between accuracy and training data volume
Solution Approach 2:
The patent introduces structural markers as an intermediary element between the raw document text and the named entity recognition process. These markers serve as mediators that guide the extraction of named entities without requiring direct training of the NER system on large annotated datasets, thereby reducing computational requirements while maintaining accuracy
2Measurement precision
If existing named entity tagging systems are used, then named entities can be extracted from documents, but the systems are computationally expensive
Solution Approach 1:
The patent segments the document processing task into distinct components: identifying structural markers (sections, lists, tables, figures) and then using these markers to guide named entity extraction. This segmentation allows the system to focus computational resources only on relevant document portions rather than processing entire documents uniformly, reducing overall computational cost while maintaining accuracy
Solution Approach 2:
The patent performs preliminary identification of structural markers before conducting named entity recognition. By pre-processing the document to locate and mark structural elements, the system prepares the data in advance, allowing the subsequent NER process to operate more efficiently with reduced computational requirements compared to systems that perform NER without such preliminary structuring
3Adaptability or versatility
If existing named entity tagging systems are used, then named entities can be extracted from English documents, but the systems are limited to specific languages and document types
Solution Approach 1:
The patent creates a universal named entity recognition approach that works across multiple languages and document types by leveraging structural markers that are common to technical documents regardless of language. The system identifies sections, numbered lists, tables, and figures using language-agnostic structural patterns, then applies named entity recognition rules that can be adapted to different languages, thereby achieving both versatility and accuracy
Data Source
AI summary
A computer implemented method for identifying one or more data elements in an input data stream formed from an input data structure using processing enabled by a processor-executable instruction can be stored on a non-transitory computerized storage medium. This stored method includes identifying a plurality of candidate named entities in one or more input document strings; identifying a plurality of enumerable markers in the one or more input document strings; associating at least one candidate named entity with at least one associated enumerable marker; associating at least one proximity string with at least one associated enumerable marker; determining a plurality of named entity descriptions based on the proximity strings; and providing the output data.