Recommendation Dialog Model Training for Diverse, Coherent Outputs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating information lack diversity and consistency, leading to repetitive and unremarkable recommendations.
Innovation Solution
A method involving splitting description information into words, inputting the sequence into a dialog generation model to obtain probability vectors, and training the model using recommendation information to enhance diversity and coherence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If existing methods for generating information are used, then the generation process is simple, but the diversity and consistency of generated recommendations deteriorate
Solution Approach 1:
The patent segments the generation process into multiple stages: obtaining probability vectors from the target object, generating candidate information through multiple paths, and selecting final recommendations. This segmentation allows the system to maintain simplicity while improving diversity by exploring multiple generation paths.
Solution Approach 2:
The patent changes key parameters including using probability vectors as input, generating multiple candidate pieces of information with different characteristics, and adjusting selection criteria to balance diversity and consistency. These parameter changes enable the system to generate more varied recommendations without significantly increasing model complexity.
2Productivity
If existing methods for generating information are used, then the processing speed is fast, but the coherence and alignment with marketing tastes deteriorates
Solution Approach 1:
The patent performs preliminary actions by first obtaining probability vectors that capture essential characteristics of the target object, then using these vectors to guide subsequent candidate generation. This preliminary step ensures coherence is established early, allowing faster generation while maintaining quality.
Solution Approach 2:
The patent implements feedback mechanisms where generated candidate information is evaluated against the probability vectors and selection criteria. This feedback loop ensures coherence with marketing tastes while maintaining efficient generation through automated selection rather than exhaustive processing.
3Device complexity
If description information is used directly without splitting, then the processing is simpler, but the quality and granularity of generated recommendations deteriorates
Solution Approach 1:
The patent applies segmentation by splitting description information into discrete probability vectors that represent different aspects or dimensions of the target object. This segmentation enables more precise and granular recommendations while keeping the processing framework relatively simple through vector-based representation.
Data Source
AI summary
The present disclosure provides a method of training an information generation model, a method of generating an information, an electronic device, and a storage medium. A specific implementation solution of the method of training the information generation model includes: splitting a description information for a target object in an information pair into at least one description word, so as to obtain a description word sequence, wherein the information pair further includes a first recommendation information; inputting the description word sequence into a dialog generation model to obtain a probability vector sequence for the target object, wherein each probability vector in the probability vector sequence includes probability values for a plurality of predetermined words; and training the dialog generation model according to the probability vector sequence and the first recommendation information, so as to obtain the information generation model.