Text Generation Model Using Logical Character Latent Representations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current text generation technologies in deep learning and natural language processing face challenges in generating coherent and logical text content, as they often rely solely on reference feature representations without effectively utilizing predetermined logical characters to improve sentence-level planning and coherence.
Innovation Solution
The method involves determining a reference feature representation of target semantic information and using predetermined logical characters to generate sentence latent representations, which are then used to create target text content, enhancing the logicality and coherence of the generated text.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If text generation relies solely on reference feature representations, then the generation process is simpler, but the logicality and coherence of generated text deteriorates
Solution Approach 1:
The patent segments the text generation process into multiple stages: determining reference feature representations from semantic information, generating sentence latent representations based on predetermined logical characters, and finally generating text content. This segmentation allows each component to focus on specific aspects (semantic meaning, logical structure, and text generation), thereby improving overall text coherence without excessively complicating the entire system.
Solution Approach 2:
The patent performs preliminary actions by first determining reference feature representations from semantic information and generating sentence latent representations before actual text generation. These preliminary steps prepare structured representations that guide the subsequent text generation process, ensuring logical consistency and coherence are built into the foundation before content creation begins.
2Reliability
If sentence-level latent representations are introduced, then text coherence improves, but model complexity increases
Solution Approach 1:
The patent introduces sentence latent representations as intermediary elements between the semantic information and the generated text. These latent representations act as mediators that capture sentence-level logical structures and semantic relationships, bridging the gap between high-level semantic intent and concrete text generation, thereby improving coherence without requiring direct complex interactions between all system components.
Solution Approach 2:
The patent changes the parameter representation by introducing latent representation vectors for sentences. Instead of directly generating text from semantic information, the system transforms semantic information into reference feature representations, then into sentence latent representations with specific dimensional vectors. This parameter transformation enables the model to capture complex logical relationships while maintaining a structured and manageable model architecture.
Data Source
AI summary
A method of generating a text, a method of training a text generation model, an electronic device, and a storage medium, which relate to a field of a computer technology, in particular to fields of deep learning and natural language processing technologies. A specific implementation solution includes: determining a reference feature representation of a target semantic information; determining, based on the reference feature representation and at least one predetermined logical character, at least one sentence latent representation respectively corresponding to the at least one predetermined logical character; and generating a target text content based on the at least one sentence latent representation.