Context-Dependent Similarity Measurement for Heterogeneous Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing similarity measurement processes for objects in heterogeneous datasets do not adequately account for contextual associations, leading to skewed similarity measures due to the presence of common feature values across different contexts.
Innovation Solution
A method that calculates context-dependent similarity scores by determining inter-context and intra-context scores for feature values, adjusting feature weights based on their co-occurrence with different and same contexts, and using these weights to generate similarity measures that consider the likelihood of objects being associated with distinct or the same contexts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional similarity functions (TF-IDF, edit distance, Jaccard distance) are used to measure object similarity, then the processing is computationally efficient and can be parallelized, but the similarity measures are skewed when common feature values appear across different contexts
Solution Approach 1:
The patent segments the similarity measurement process into context-dependent components. It divides the feature space by introducing context identifiers that partition objects into different context groups. This segmentation allows the system to evaluate feature similarity separately within each context, preventing cross-context contamination where common features from different contexts artificially inflate similarity scores.
Solution Approach 2:
The patent applies local quality by making the similarity measurement adaptive to local context characteristics. Instead of using a uniform similarity function across all objects, it adjusts the similarity calculation based on the specific context of each object pair. Features are weighted differently depending on their contextual relevance, allowing the same feature to contribute differently to similarity scores based on the objects' contexts.
2Measurement precision
If context information is incorporated into similarity measurements to account for contextual associations, then the accuracy of similarity scores improves, but the computational complexity and data processing requirements increase
Solution Approach 1:
The patent performs preliminary action by pre-computing and storing context information alongside object features. Context identifiers and context-based feature weights are calculated in advance and cached, so that during actual similarity queries, the system can directly use these pre-computed values without performing expensive re-analysis. This preliminary preparation significantly reduces the computational burden during runtime similarity measurements.
3Ease of manufacture
If feature values that are common across multiple contexts are given equal weight in similarity calculations, then the processing is simple and straightforward, but the similarity measures become skewed and less reliable
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting feature weights based on contextual parameters. Instead of using fixed weights for all features, it modifies the weight of each feature according to the context in which it appears. Common features that appear across multiple contexts receive lower weights or are adjusted based on their distribution pattern, while context-specific features receive higher weights, making the similarity measurement more reliable.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for determining similarity measures for objects in a dataset that include contextual associations of the objects with contexts. In one aspect, a method includes calculating a similarity measure for any two objects that include a common feature f based, in part, on the likelihood that the two object representations in the dataset that both include f will we associated with distinct contexts, and the likelihood that the two objects in the dataset that both include f will be associated with the same context.