Recommendation Message Generation Using Seed Attributes and Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating recommendation messages, whether through neural networks or templates, fail to accurately reflect the characteristics of to-be-recommended content, resulting in low conversion rates.
Innovation Solution
A method involving obtaining a content description, predicting a seed attribute, performing information retrieval to obtain supplementary information, populating a prompt template with the content and supplementary information, and generating a recommendation message using a first large-scale pre-trained language model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If neural network models or fixed templates are used to generate recommendation messages, then the generation process is simple and fast, but the accuracy and conversion rate of recommendation messages are low
Solution Approach 1:
The patent segments the recommendation message generation process into multiple independent modules: content analysis module, attribute prediction module, information retrieval module, template selection module, and message generation module. Each module handles a specific task, allowing the system to achieve high accuracy through specialized processing while maintaining manageable complexity through modular architecture.
Solution Approach 2:
The patent introduces an information database as an intermediary between the content analysis and message generation processes. This database stores pre-collected supplementary information about various content types, which is retrieved based on predicted attributes and used to enhance the recommendation messages. This intermediary component enables accurate, context-rich message generation without requiring the entire system to be overly complex.
2Adaptability or versatility
If fixed templates are applied regardless of content type, then the generation process is efficient, but the recommendation messages fail to reflect actual content characteristics
Solution Approach 1:
The patent implements a dynamic template selection mechanism where the template is not fixed but is chosen based on the predicted attributes of the content. The system dynamically adjusts the template structure and content based on the specific characteristics of each recommendation item, ensuring adaptability to different content types while maintaining generation efficiency through pre-defined template patterns.
Solution Approach 2:
The patent changes the parameters of the recommendation message generation process based on content attributes. Different content types (news, products, services, etc.) trigger different generation parameters including template selection, information retrieval queries, and message structure. This parameter-based approach enables the system to adapt to various content characteristics while maintaining a systematic and efficient generation process.
3Measurement precision
If more supplementary information is retrieved and processed, then the accuracy of recommendation messages improves, but the processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary actions by pre-collecting and organizing supplementary information about various content types in an information database before the actual recommendation generation occurs. This pre-processing stage includes gathering attributes, descriptions, and contextual information for different content categories. When a recommendation is needed, the system only needs to retrieve pre-processed information based on predicted attributes, significantly reducing the time and computational resources required during the actual generation process while maintaining high accuracy.
Data Source
AI summary
A method for generating a recommendation message includes: obtaining a recommendation request for content, the recommendation request containing a content description for recommendation; predicting a seed attribute of the content based on the content description; performing information retrieval in an information database based on the content description and the seed attribute to obtain supplementary information corresponding to the content description and the seed attribute; populating a prompt template with the content description and the supplementary information to obtain a query message; and generating, based on the query message, a recommendation message of the content for recommendation using a first large-scale pre-trained language model.