Hierarchical Data Processor Tester for Cognitive Systems

The patent transforms the testing approach by changing the parameter of string comparison from exact matching to similarity-based matching. It introduces a hierarchical set of similarity operators (e.g., Levenshtein distance, Jaccard similarity, cosine similarity) that compute similarity scores between output strings and expected strings, allowing the system to identify correct translations even when character sequences differ slightly. This parameter change resolves the contradiction by maintaining measurement precision while reducing false negatives.

The patent implements a dynamic, multi-level testing framework where the strictness of comparison is adjusted based on the testing context. The hierarchical similarity operators provide multiple levels of comparison strictness, and the system can dynamically select appropriate operators or adjust similarity thresholds based on the specific testing scenario. This dynamic approach allows the system to be stringent when needed while being tolerant of legitimate variations, thereby improving reliability without sacrificing precision.

The patent replaces the mechanical human testing process with an automated computer-based system that uses algorithms to perform similarity comparisons. Instead of human reviewers manually evaluating each output, the system automatically applies hierarchical similarity operators to compute match scores and determine accuracy. This substitution maintains high reliability through sophisticated comparison logic while dramatically improving productivity by processing numerous test cases quickly and at low cost.

The patent introduces similarity score computation as an intermediary between the raw output comparison and the final accuracy determination. Rather than directly comparing outputs or relying on human judgment, the system uses similarity operators as intermediaries that quantify the degree of match between actual and expected outputs. This intermediary layer enables automated systems to capture the nuanced judgment that humans would make, achieving both reliability and productivity.

The patent segments the similarity comparison process into a hierarchical structure with multiple distinct operators, each handling specific types of string variations. The hierarchy typically progresses from exact matching (most deterministic) to increasingly permissive similarity measures (more probabilistic). This segmentation allows the system to tackle different aspects of string comparison separately, reducing false negatives by ensuring that each type of variation is handled by the appropriate operator, while keeping the overall system manageable through modular design.

The patent creates a composite testing system that combines multiple similarity operators with different characteristics into a unified hierarchical framework. Each operator can be viewed as a different 'material' with specific properties (e.g., strictness, computational cost, types of variations handled). By compositing these operators in a hierarchical structure, the system achieves robust accuracy determination that leverages the strengths of each individual operator while mitigating their individual weaknesses, thereby reducing false negatives without overwhelming complexity.