Risk-Aware Entity Linking Using Semantic and Quality Filters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current entity resolution techniques in computer systems face challenges in accurately linking entities across heterogeneous data sources, often resulting in low-quality or high-risk entity links, which can lead to unnecessary and irrelevant notifications, and increased load on content delivery systems.
Innovation Solution
The system computes risk data and semantic similarity data to determine whether to link entities from multiple digital data sources, using digital models to analyze features of content items and entities, and applying programmable rules to filter out low-quality links, thereby improving the accuracy and efficiency of entity linking operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If entity resolution is performed using natural language processing techniques to match entities across data sources, then entity linking capability is established, but link quality deteriorates resulting in low-quality or high-risk entity links
Solution Approach 1:
The patent introduces risk data as an intermediary factor between semantic similarity matching and entity link creation. The risk evaluation component analyzes multiple features (entity features, content features, link features) and produces risk data that mediates the final linking decision, preventing low-quality links even when semantic similarity is high
Solution Approach 2:
The system changes the parameter space by adding risk assessment dimensions beyond traditional semantic similarity. Instead of relying solely on matching accuracy, the system evaluates entity quality, content quality, and link quality parameters, transforming the linking decision from a single-parameter problem to a multi-parameter optimization problem
2Productivity
If comprehensive entity matching is performed across heterogeneous data sources, then entity linking coverage is improved, but system load increases due to unnecessary notifications and processing
Solution Approach 1:
The system performs preliminary risk evaluation before creating entity links and notifications. By assessing entity quality, content quality, and link quality in advance, the system filters out low-quality candidates before they consume notification and delivery resources, reducing overall system load while maintaining comprehensive coverage
3Speed
If entity linking is performed without risk assessment, then processing speed is maintained, but link accuracy deteriorates resulting in irrelevant notifications
Solution Approach 1:
The entity linking process is segmented into distinct evaluation stages: semantic similarity matching, entity quality assessment, content quality assessment, link quality assessment, and final linking decision. This segmentation allows parallel processing of multiple features while maintaining comprehensive accuracy evaluation without sacrificing overall processing speed
Data Source
AI summary
In an embodiment, the disclosed technologies include identifying a content item of a first digital data source as a candidate for linking with a target entity of a second digital data source by matching a candidate entity mentioned in the content item to the target entity in accordance with semantic similarity data computed between the candidate entity and the target entity; inputting at least one feature of the content item and at least one feature of the target entity to a set of digital models that analyze the at least one feature of the content item and the at least one feature of the target entity and determine and output qualitative data; based on the qualitative data, determining link risk data; based on the link risk data and the semantic similarity data, and determining whether to generate a link between the content item and the target entity.