Alphanumeric String Similarity Analysis via Cosine Self-Join

Solution Overview

Problem

Manual analysis of alphanumeric strings to find related groups is time-consuming and prone to errors, especially when dealing with large datasets, as it requires manual comparison and detection of similarity relationships.

Innovation Solution

A system and method utilizing a back-end application computer server to automatically analyze alphanumeric strings by computing cosine similarity scores, constructing a two-column result table via self-join, and outputting similarity indications, allowing for fast and accurate identification of related strings.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual review is used to analyze alphanumeric strings for similarity relationships, then analysis accuracy can be maintained through human judgment, but the process becomes time-consuming and expensive especially when substantial numbers of strings need to be analyzed

Engineering Contradiction:

Improvesimilarity detection accuracyVSAvoidanalysis time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces manual mechanical review with an automated computerized system that uses cosine similarity algorithms to analyze alphanumeric strings. The system processes strings through mathematical computations (converting text to vectors and calculating cosine similarity scores) to automatically identify similar strings, eliminating the need for time-consuming manual comparison while maintaining detection accuracy through objective mathematical metrics.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Adaptability or versatility

If manual comparison is used to evaluate alphanumeric strings, then flexibility in detecting various levels of similarities can be achieved, but the complexity of appropriately responding to different similarity levels increases

Engineering Contradiction:

Improvesimilarity level detection flexibilityVSAvoidresponse complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent uses cosine similarity scores as a continuous parameter that naturally captures various levels of string similarity. By computing the cosine angle between vector representations of strings, the system produces a numerical score that reflects the degree of similarity. This allows the system to handle multiple similarity levels automatically through a single mathematical parameter, simplifying the response complexity while maintaining adaptability to detect different degrees of string relationships.

Inventive Principle:

Principle #35Parameter changes

3Productivity

If automated methods are used to analyze alphanumeric strings, then analysis speed and productivity improve, but the ability to accurately detect similarity relationships and interpret various levels of similarity may be compromised

Engineering Contradiction:

Improveanalysis speedVSAvoidsimilarity detection accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent replaces manual mechanical review with an automated computerized system that uses cosine similarity algorithms to analyze alphanumeric strings. The system processes strings through mathematical computations (converting text to vectors and calculating cosine similarity scores) to automatically identify similar strings, eliminating the need for time-consuming manual comparison while maintaining detection accuracy through objective mathematical metrics.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent uses cosine similarity scores as a continuous parameter that naturally captures various levels of string similarity. By computing the cosine angle between vector representations of strings, the system produces a numerical score that reflects the degree of similarity. This allows the system to handle multiple similarity levels automatically through a single mathematical parameter, simplifying the response complexity while maintaining adaptability to detect different degrees of string relationships.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20220253471A1Systems and methods to evaluate a column of text strings to find relationships

Publication Date: 2022.08.11 HARTFORD FIRE INSURANCE CO

AI summary

Embodiments may be associated with an alphanumeric string similarity analysis system implemented via a back-end application computer server. The computer server may receive information about electronic records to be analyzed, wherein each electronic record is associated with enterprise data and includes an electronic record identifier and an alphanumeric string. The computer server may then store the alphanumeric strings in a single column and compute a length of each alphanumeric string in the single column. A two-column result table may be constructed via a self-join on the single column, with shorter strings being stored in a first column of the result table. The result table may be automatically analyzed using cosine similarity to generate a similarity score for an alphanumeric string in the first column and a corresponding string in a second column of the result table. It may then be arranged for indications of the similarity scores to be output.