Prompt Processing by Removing Irrelevant Information Elements
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing Generative Artificial Intelligence (GAI) systems consume excessive power and resources due to processing of text prompts with irrelevant Information Elements (IEs, leading to increased computational cost and power requirements.
Innovation Solution
A method and system for identifying and removing irrelevant IEs from text prompts by determining semantic and structural proximity values, internal consistency, and applying predefined criteria to generate a revised prompt, which is then submitted to the GAI system for processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If the complete text prompt with all Information Elements is submitted to the GAI system, then the system can process comprehensive information, but power consumption and computational resources increase excessively
Solution Approach 1:
The patent extracts and removes irrelevant Information Elements from the text prompt before submission to the GAI system. The system identifies and eliminates redundant, duplicate, and non-essential IEs while preserving critical information elements that are necessary for accurate response generation, thereby reducing power consumption without significant loss of information completeness
Solution Approach 2:
The patent applies different processing treatments to different Information Elements based on their relevance and importance. Critical IEs are preserved with high fidelity, while less important IEs are selectively removed or modified. This localized quality approach ensures that power consumption is reduced by removing only the non-essential portions while maintaining the integrity of essential information
2Use of energy by moving object
If irrelevant Information Elements are removed from the text prompt, then power consumption decreases, but the risk of losing relevant information increases
Solution Approach 1:
The patent employs feedback mechanisms where the system evaluates the impact of removing each Information Element on the overall prompt quality and response accuracy. The removal process is iterative, with the system monitoring whether essential information is being preserved while achieving power consumption reduction goals, allowing dynamic adjustment of the filtering criteria
Solution Approach 2:
The patent performs preliminary analysis and classification of Information Elements before the actual removal process. By pre-identifying critical IEs that must be preserved and categorizing removal candidates, the system ensures that relevant information is protected from accidental removal while enabling aggressive filtering of irrelevant content to maximize power savings
3Productivity
If the text prompt is processed without filtering, then all information is available for processing, but computational resources and processing time are wasted on irrelevant elements
Solution Approach 1:
The patent segments the text prompt into discrete Information Elements and processes them individually through filtering criteria. This segmentation allows the system to efficiently identify and remove irrelevant IEs in parallel, reducing the overall computational burden and processing time compared to analyzing the complete prompt as a single unit
4Reliability
If the text prompt includes redundant and duplicate Information Elements, then comprehensive coverage is achieved, but the prompt length and processing overhead increase
Solution Approach 1:
The patent discards redundant and duplicate Information Elements that do not add value to the response quality. By identifying and removing duplicate IEs while preserving the essential information content, the system reduces prompt complexity and processing overhead while maintaining response accuracy through the recovery and retention of unique, non-redundant information elements
Data Source
AI summary
Methods, systems, and computer-readable storage media for processing a text prompt including a set of text Information Elements (IEs). Original instruction semantic IEs and original contextual IEs are identified within the text IEs. Some of the original instruction semantic IEs are identified for removal from the text prompt, based on semantic proximity values and internal consistency values of the instruction semantic IEs relative to first predefined criteria, while leaving surviving instruction semantic IEs. Similarly, some of the original contextual IEs are identified for removal from the text prompt due to weak connections with the surviving instruction semantic IEs and other of the contextual IEs based on second predefined criteria, while leaving surviving contextual IEs. Further, a revised text prompt corresponding to the surviving instruction semantic IEs and the surviving contextual IEs is generated and submitted as a query to a GAI system programmed to answer the query.