Entity Linking via Confidence Score Iteration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current technologies face challenges in accurately linking documents with relevant entities and actions, particularly in disambiguating mentions of entities within documents and associating them with appropriate actions, which hinders the ability to provide users with relevant actions and applications during online searches.
Innovation Solution
The method involves identifying mentions of entities within documents, updating confidence scores based on relationships between candidate entities, and selecting entities with scores above a threshold to rank their relevance, allowing for the identification of associated actions and applications, which are then linked within search results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional entity linking methods are used without considering relationships between candidate entities, then the processing speed is faster, but the accuracy of entity linking deteriorates
Solution Approach 1:
The system updates confidence scores iteratively by incorporating relationship information from candidate entities. The confidence score for each mention-entity pair is refined through feedback loops that consider relationships between candidate entities, thereby improving entity linking accuracy while managing processing complexity through structured iterations.
Solution Approach 2:
The patent introduces confidence scores as an intermediary mechanism to bridge the gap between simple entity matching and complex relationship analysis. By using confidence scores that are updated based on entity relationships, the system mediates between processing speed and accuracy, allowing efficient yet accurate entity linking.
2Measurement precision
If entity relationships are analyzed to improve linking accuracy, then the precision of entity identification is improved, but the time required for processing increases
Solution Approach 1:
The system performs preliminary identification of candidate entities and pre-calculates relationship information before final entity selection. By preparing confidence scores and relationship data in advance, the system reduces the time required for final entity identification while maintaining high precision through relationship-based scoring.
Solution Approach 2:
The patent dynamically adjusts confidence score parameters based on relationship strength and entity context. By changing the weighting parameters of entity relationships during processing, the system optimizes the balance between analysis depth and processing time, achieving high precision without excessive time consumption.
3Reliability
If multiple candidate entities are considered for each mention, then the accuracy of entity selection is improved, but the complexity of the selection process increases
Solution Approach 1:
The patent segments the entity selection process into distinct stages: candidate identification, confidence score initialization, relationship-based score updating, and final selection. By dividing the complex selection process into manageable segments, the system maintains high reliability through systematic evaluation while reducing overall process complexity through structured organization.
Solution Approach 2:
The system uses confidence score parameters to manage the complexity of evaluating multiple candidate entities. By adjusting and updating these parameters based on entity relationships, the system systematically narrows down candidates while maintaining reliable selection, transforming a complex multi-entity evaluation into a parameter-driven decision process.
Data Source
AI summary
Individuals often perform search queries with the intent of performing a particular action based on the search results. In an embodiment, mentions are identified in a document and candidate entities for the mention are identified. Confidence scores for mention-entity pairs are determined based on relationships between candidate entities, and entities are selected based on the scores. Selected entities are ranked indicating their relevance to a topic of the document. Actions associated with the selected entities are identified. Based on an intent of a search query that returned the document in search results, relevant actions are selected and links are presented within the search results to applications that perform the actions.