LLM Prompt Generation for Flexible Content Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional content filtering techniques rely on predetermined rules, making it difficult to flexibly filter content according to a user's specific requests.
Innovation Solution
An information processing device that acquires reader information and content information, generates a prompt based on this data, and inputs the prompt into a large language model to produce edited content that reflects the user's preferences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional filtering techniques use predetermined rules, then filtering can be performed with simple processing, but the system cannot flexibly adapt to user-specific requests
Solution Approach 1:
The patent introduces a prompt generation unit as an intermediary component that bridges user information and the large language model. This unit synthesizes user profiles, content information, and filtering rules into structured prompts, enabling flexible adaptation without directly complicating the core filtering system. The intermediary handles the complexity of integrating multiple data sources while presenting a simplified interface to the LLM.
Solution Approach 2:
The system employs a large language model that autonomously generates filtered content based on input prompts without requiring manual rule configuration for each filtering scenario. The LLM self-adapts to different user requests by processing the generated prompts, eliminating the need for complex manual rule management while maintaining high adaptability to user-specific needs.
2Adaptability or versatility
If conventional filtering uses predetermined rules, then the system structure remains simple, but it cannot reflect user information in content editing
Solution Approach 1:
The system performs preliminary synthesis of user information, content characteristics, and filtering requirements into structured prompts before submitting them to the large language model. This preliminary action organizes complex user data into a format that the LLM can efficiently process, enabling the system to reflect user information accurately while managing structural complexity through advance preparation.
Solution Approach 2:
The patent dynamically adjusts the parameters and structure of prompts based on user profiles and content types. By changing the formulation, detail level, and focus of prompts according to specific user characteristics, the system adapts to different user information requirements without requiring a fundamentally different system structure, thus balancing adaptability with manageable complexity.
3Manufacturing precision
If flexible content editing is implemented using large language models, then user preferences can be accurately reflected, but processing time and computational resources increase
Solution Approach 1:
The system applies filtering and editing actions selectively based on the severity and type of content issues detected. For minor adjustments, it performs partial editing only on specific segments rather than complete content regeneration. This approach maintains high accuracy where needed while reducing unnecessary processing time for content that requires minimal or no modification.
Solution Approach 2:
The prompt generation unit prepares optimized, concise prompts that contain only the essential information needed for accurate content editing. By pre-structuring the input data to highlight critical user preferences and content issues, the system reduces the computational burden on the large language model, thereby decreasing processing time while maintaining editing accuracy.
Data Source
AI summary
The acquisition means acquires reader information and content information. The prompt generation means generates a prompt based on the reader information and the content information. The content editing means inputs the prompt into a large language model and acquire an output from the large language model as edited content.