Text Generation Model With Sub-Text Evaluation for Literary Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing deep learning models struggle to generate high-quality literary works such as poems and Ci, couplets, and fictions, as they fail to ensure the semantic and literary features meet preset conditions.
Innovation Solution
A text generation method and apparatus that utilizes a text information generation model, a first evaluation model, and a second evaluation model to determine and splice sub-text information based on semantic and literary features, ensuring the final text meets preset literary and semantic conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a deep learning model is used for text generation, then the generation speed and automation are improved, but the quality and literary features of the generated text deteriorate
Solution Approach 1:
The patent segments the text generation process into multiple sequential steps, where each step generates a portion of the text and evaluates it against literary features. This segmentation allows the system to maintain high generation speed while ensuring quality at each stage through targeted evaluation of semantic coherence, rhythmic patterns, and literary characteristics.
Solution Approach 2:
The patent implements a feedback mechanism where the deep learning model's generated text is evaluated against preset literary features and semantic requirements. The evaluation results are fed back to guide subsequent generation steps, allowing the system to automatically adjust and improve text quality while maintaining efficient generation throughput.
2Manufacturing precision
If the text generation model focuses on literary features, then the text quality is improved, but the generation complexity increases
Solution Approach 1:
The patent divides the complex task of generating high-quality literary text into multiple simpler sub-tasks, each focusing on specific literary features such as semantics, rhythm, or rhyme. This segmentation reduces the complexity of each individual generation step while collectively achieving high overall text quality.
Solution Approach 2:
The patent performs preliminary evaluation and guidance before final text generation by establishing preset literary features and semantic requirements. This preliminary action prepares the generation process in advance, reducing the complexity of real-time decision-making during text generation while ensuring quality outcomes.
3Reliability
If multiple evaluation models are used to ensure text quality, then the text reliability is improved, but the processing time increases
Solution Approach 1:
The patent segments the evaluation process into multiple specialized evaluation models, each focusing on specific aspects of text quality such as semantic coherence, rhythmic patterns, or literary features. This segmentation allows parallel or sequential evaluation of different aspects without requiring a single comprehensive evaluation, thereby maintaining reliability while managing processing time efficiently.
Solution Approach 2:
The patent applies partial evaluation at each generation step rather than complete re-evaluation of the entire text. This partial action approach ensures quality reliability by evaluating only the necessary portions of text at each step, reducing the total processing time while maintaining comprehensive quality assurance through cumulative evaluation.
Data Source
AI summary
The present disclosure provides a text generation method and apparatus, a device, and a storage medium. The text generation method includes: obtaining text information to be processed; and inputting the text information to be processed to a text generation model to obtain target text information corresponding to the text information to be processed, the target text information is determined by splicing the N pieces of sub-text information in a preset order, with a literary feature of the target text information meeting a preset condition; first sub-text information of the target text information is determined by the text generation model according to a semantic feature of the text information to be processed; and i-th sub-text information of the target text information is determined by the text generation model according to literary features of the first sub-text information to (i−1)-th sub-text information.