AI Content Generation via Requirements Driven Outlines

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Transformer-based language models struggle to automatically generate text that satisfies complex requirements, particularly when dealing with multi-variate unstructured text, as they require seed material that is relevant and structured.

Innovation Solution

A system that ingests multi-variate data with unstructured content, trains a machine-learning model using deep learning and natural language understanding techniques to structure and label sections, and generates Requirements Driven Outlines (RDOs) to guide the generation of semantically relevant content.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If transformer-based language models are used to generate text, then text generation capability is provided, but the models struggle to understand multi-variate unstructured text and require structured seed material

Engineering Contradiction:

Improvecapability to handle unstructured textVSAvoidaccuracy of requirement satisfaction

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The patent introduces an intermediary NLP processing layer between the unstructured text input and the language model. This intermediary component (the NLP model) analyzes, segments, and structures the unstructured text into formatted representations that the language model can reliably process, thereby resolving the contradiction between handling unstructured text and maintaining generation accuracy

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent segments the unstructured text into discrete sections or segments through NLP processing. By breaking down the complex unstructured text into manageable segments that are individually analyzed and labeled, the system enables the language model to process the information more reliably while maintaining versatility in handling various unstructured formats

Inventive Principle:

Principle #1Segmentation

2Measurement precision

If human annotators are used to label sections of documents, then accurate labeling is achieved, but the process is time-consuming and labor-intensive

Engineering Contradiction:

Improvelabeling accuracyVSAvoidtime required for annotation

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by using an NLP model to pre-label and structure document sections before human review. The NLP model performs the initial segmentation and labeling task, reducing the amount of manual work required while maintaining high accuracy through subsequent human validation and feedback mechanisms

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent implements feedback loops where human annotators review and correct NLP model predictions, and this feedback is used to refine and retrain the NLP model. This iterative feedback process improves labeling accuracy over time while reducing the overall time investment compared to purely manual annotation

Inventive Principle:

Principle #23Feedback

3Productivity

If machine-learning models are trained on subsets of documents, then training efficiency is improved, but the model may not capture all nuances of the full dataset

Engineering Contradiction:

Improvetraining speedVSAvoidcompleteness of knowledge representation

Core Design Contradiction:

ProductivityVSManufacturing precision

Solution Approach 1:

The patent applies dynamics by implementing a dynamic training approach where the model is trained on subsets of data and then continuously refined through feedback from human annotators and new data inputs. This allows the model to maintain training efficiency while progressively improving its knowledge representation completeness through ongoing learning and adaptation

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS12299404B2Computer-generated content based on text classification, semantic relevance, and activation of deep learning large language models

Publication Date: 2025.05.13 ROHIRRIM INC

AI summary

The disclosure relates to systems and methods of automatically generating unique content including natural language text based on a corpus of previously generated response documents and discrete requirements defined in a requirements specification. The system may use generative stitching that includes multi-layer processes that execute to influence the generation of unique content including natural language text through an artificial intelligence (AI) language transformer model trained to output the content based on previously written material that is semantically relevant to the discrete requirements and is weighted against labeled attributes. The labeled attributes may determine the influence asserted against the language transformer, thereby generating unique on-target content that may be combined to create a computer-generated response document.