Automated Ontology Mapping for NER False Positives
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current named entity recognition (NER) processes face challenges in linking synonyms, mapping to ontological concepts, and dealing with false positives, which are computationally intensive and laborious.
Innovation Solution
The proposed solution involves a method that learns new synonyms for a given concept without manual curation, relates raw NER outputs to ontological concepts, accounts for false positives, aggregates NER results from different approaches, and presents knowledge graphs in user-adjustable ways.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual curation techniques are used to link synonyms and map to ontological concepts, then mapping precision is improved, but productivity deteriorates due to labor intensity and time consumption
Solution Approach 1:
The patent replaces manual curation processes with automated machine learning models that learn to map NER outputs to ontological concepts. The system uses neural networks to automatically associate entity mentions with ontology terms, eliminating the need for human curators to manually create and maintain synonym links while achieving comparable or superior mapping precision.
Solution Approach 2:
The system enables automatic self-learning where the mapping model is trained on annotated data and then autonomously performs synonym linking and ontology mapping without human intervention. The model continuously improves by learning from labeled examples and can independently handle new entity-ontology relationships that arise in the data.
2Productivity
If computational methods are used to handle large-scale unstructured data, then productivity is improved, but reliability deteriorates due to false positives in NER processes
Solution Approach 1:
The patent implements feedback mechanisms where the system learns from annotated data and corrects false positives by training on labeled examples that include correction information. The model uses feedback loops to refine its predictions, gradually improving reliability as it processes more data and learns from its errors through supervised learning.
Solution Approach 2:
The system performs beforehand validation by cross-referencing NER outputs against multiple data sources and using confidence thresholds to filter out low-quality predictions before they enter downstream processing. This pre-filtering approach cushions against false positives by eliminating unlikely candidates before they cause problems.
3Reliability
If multiple NER approaches are aggregated to improve reliability, then reliability is improved, but device complexity increases
Solution Approach 1:
The patent merges multiple NER approaches by integrating different algorithms and models into a unified system that processes entity recognition through multiple stages. The system combines rule-based extraction, machine learning models, and ontology-based validation into a single coordinated pipeline that leverages the strengths of each approach while filtering out their weaknesses through integration.
Data Source
AI summary
This disclosure enables various technologies that can (1) learn new synonyms for a given concept without manual curation techniques, (2) relate (e.g., map) some, many, most, or all raw named entity recognition outputs (e.g., “United States”, “United States of America”) to ontological concepts (e.g., ISO-3166 country code: “USA”), (3) account for false positives from a prior named entity recognition process, or (4) aggregate some, many, most, or all named entity recognition results from machine learning or rules based approaches to provide a best of breed hybrid approach (e.g., synergistic effect).