Entity Attribute Inference via Partial-to-Partial Similarity Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in maintaining accurate databases of attribute values for entities, often resulting in missing or ambiguous information due to incomplete or unextracted data from web content, which hampers effective web ranking for online search features.
Innovation Solution
A system that infers missing attribute values by combining partial information from multiple heterogeneous data sources using a partial-to-partial similarity graph, normalizing and aggregating data to generate web ranking signals, allowing for the inference of attribute values based on related entities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If databases are created by mining crawled Web content using templates, then attribute values can be extracted from available sources, but missing or ambiguous attribute values occur when information is not present on the Web or cannot be successfully extracted
Solution Approach 1:
The patent introduces an inference engine as an intermediary component that bridges the gap between incomplete extracted data and complete entity attribute databases. This inference engine uses probabilistic reasoning and relationships between entities to fill in missing attribute values, effectively mediating between the extraction system's limitations and the need for complete information.
Solution Approach 2:
The system implements feedback mechanisms where inference results are fed back into the database to improve future extractions. The inference engine analyzes patterns in existing data and uses this feedback to refine extraction templates and improve the identification of missing attribute values in subsequent processing cycles.
2Loss of information
If multiple heterogeneous data sources are integrated to improve attribute value completeness, then more information can be obtained, but data normalization and aggregation complexity increases
Solution Approach 1:
The patent creates a universal data model that can accommodate multiple heterogeneous data sources through a common schema. This universal representation layer translates diverse input formats into a standardized internal structure, enabling the system to process various data sources without requiring separate handling logic for each source type.
Solution Approach 2:
The system dynamically adjusts parameters such as confidence thresholds and weighting factors based on data source reliability and attribute importance. By changing these parameters adaptively, the system optimizes the integration of heterogeneous sources without requiring complex manual configuration for each new data source.
3Reliability
If entity attribute databases are maintained for web ranking, then search results can be improved, but the databases suffer from incomplete information that hampers effective web ranking
Solution Approach 1:
The system performs preliminary inference of missing attribute values before the web ranking process. By proactively filling in missing information using relationships between entities and probabilistic reasoning, the database is pre-prepared with more complete information, thereby improving ranking accuracy without requiring complete information during the actual ranking operation.
Solution Approach 2:
The patent segments the attribute completion problem into manageable parts by processing entities in groups based on their relationships. Instead of attempting to complete all attributes for all entities simultaneously, the system divides the work into segments that can be processed independently and then combined, improving both efficiency and accuracy.
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
Systems, methods, and computer-readable storage media are provided for inferring missing or ambiguous attribute values for entities based on partial information about such entities and/or information about other similar entities as extracted from multiple information sources for generating Web ranking signals for online search. A plurality of heterogeneous input data sources are ingested and combined to produce output data having information content that is more than the sum of its parts. A generic platform is provided where multiple data sources having information content related to entity attributes can be plugged-in without additional changes being necessary to the platform. This generic plugin model for extracting and inferring entity attribute values makes it easy to leverage new data sources as they become available to improve the final inferred attribute data.