Requirements-to-Code Similarity Matrix Using Semantic Vectors

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional application development tools lack the ability to reduce the gap between intended requirements and the final delivered product, requiring significant manual effort and lacking automation for generating a similarity matrix between intended requirements context data and source code context data.

Innovation Solution

A language-agnostic application development module that automatically generates a similarity matrix/score between intended requirements context data and source code context data, providing code suggestions, improving code accuracy, and estimating development completion by converting data into semantic context vectors and using a cosine similarity algorithm.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If conventional application development tools are used, then development can proceed with existing manual processes, but the gap between intended requirements and final delivered product cannot be reduced

Engineering Contradiction:

Improvealignment between requirements and delivered productVSAvoidautomation for generating similarity matrix

Core Design Contradiction:

Manufacturing precisionVSExtent of automation

Solution Approach 1:

The patent introduces an intermediary system comprising a requirements parser, code analyzer, and similarity computation module that mediates between requirements data and source code. This intermediary automatically generates a similarity matrix by comparing semantic representations of requirements and code, thereby reducing the gap between intended requirements and delivered product without requiring full manual intervention.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces manual mechanical processes of requirement analysis and code verification with an automated computational system. The system uses natural language processing to parse requirements, abstract syntax tree generation to analyze code, and cosine similarity computation to automatically generate a similarity matrix, substituting manual effort with automated mechanical-computational processes.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If manual processes are used for requirement analysis and code verification, then detailed attention can be given, but significant manual effort is required

Engineering Contradiction:

Improveaccuracy of requirement-code alignmentVSAvoiddevelopment efficiency

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The system enables self-service by automatically performing requirement analysis, code parsing, semantic comparison, and similarity matrix generation without requiring manual intervention. The automated pipeline processes requirements data and source code independently, computing alignment metrics and providing feedback to developers, thereby maintaining measurement precision while significantly improving productivity.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent performs preliminary actions by automatically generating the similarity matrix early in the development process, before final code delivery. This preliminary automated analysis provides upfront feedback on requirement-code alignment, allowing developers to correct discrepancies before completion, thereby maintaining high accuracy while improving overall development efficiency.

Inventive Principle:

Principle #10Preliminary action

3Loss of information

If no automated similarity assessment is implemented, then development process remains simple, but real-time feedback on development progress is not provided

Engineering Contradiction:

Improvefeedback on development progressVSAvoidcomplexity of development tooling

Core Design Contradiction:

Loss of informationVSDevice complexity

Solution Approach 1:

The patent implements a feedback mechanism where the automatically generated similarity matrix provides real-time quantitative feedback on the alignment between requirements and source code. This feedback is computed by comparing semantic representations using cosine similarity and presented to developers, enabling them to track development progress and ensure requirement adherence without significantly increasing tooling complexity.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12461714B2System and method for generating a similarity matrix/score between intended requirements context data and source code context data

Publication Date: 2025.11.04 JPMORGAN CHASE BANK NA

AI summary

Various methods, apparatuses/systems, and media for developing an application are disclosed. A processor converts requirements data into a first semantic context data, the requirements data describing intended tasks required for developing an application; converts the first semantic context data into a first semantic context vector; accesses a database that stores source code corresponding to implementation of the intended tasks required for developing the application; converts the source code into a second semantic context data; converts the second semantic context data into a second semantic context vector; compares the first semantic context vector and the second semantic context vector; automatically generates, in response to comparing, a similarity score that indicates how much the source code and the requirements data are in line with each other; and executes development of the application when it is determined that the similarity score is equal to or more than a predetermined threshold value.