Missing Fact Estimation in Knowledge Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Relational models of knowledge, such as graph-based data stores, often lack complete information, with missing or incorrect facts, making it difficult to provide accurate answers to search queries, as they rely on incomplete data.
Innovation Solution
A computer system that uses a joint distribution model to infer missing facts by analyzing related features in the data graph, generating estimates and providing human-readable explanations based on influential features, allowing for the selection and storage of key features that contribute to these estimates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If a graph-based data store is used to provide search answers, then a significant amount of facts can be represented, but tens of millions of entities may be missing or have incorrect facts
Solution Approach 1:
The system pre-computes and stores feature vectors for all entities in the graph, capturing relationships and attributes beforehand. This preliminary action enables rapid inference of missing facts during queries without requiring complete pre-existing information for every entity
Solution Approach 2:
The system introduces feature vectors as intermediary representations that mediate between the graph structure and inference operations. These feature vectors encode entity characteristics and relationships, serving as a bridge that enables statistical inference to fill missing information
2Measurement precision
If estimates for missing facts are generated using multiple features, then accuracy improves, but the complexity of determining influential features increases
Solution Approach 1:
The system computes feature vectors that include more features than strictly necessary, then uses statistical methods to identify the most influential subset. This partial action approach computes comprehensive features upfront but only utilizes the most significant ones for inference, balancing accuracy with computational efficiency
Solution Approach 2:
The system transforms raw graph data into feature vector representations, changing the parameter space from structured graph relationships to continuous feature values. This transformation enables the application of statistical inference techniques and facilitates the identification of influential features through quantitative analysis
3Reliability
If comprehensive feature analysis is performed to generate explanations, then user trust improves, but the time required to process queries increases
Solution Approach 1:
The system pre-computes feature vectors and stores them in association with each entity, performing the computationally intensive feature extraction beforehand. During query processing, the system only needs to retrieve pre-computed features and generate explanations, significantly reducing query response time while maintaining comprehensive analysis capabilities
Data Source
AI summary
Systems and methods are disclosed for providing an explanation of an estimate for information missing from a data graph. An example method may include receiving a query that requests information for a first entity and receiving an estimate for the information, the estimate being based on a plurality of features of a joint distribution model. The method may include determining respective contribution scores for the plurality of features, selecting a quantity of the features with highest contribution scores, generating, using the selected quantity of features, an explanation for the estimate; and providing the explanation and the estimate as part of a search result for the query.