Self-Encoding Neural Network Text Generation via Hidden Feature Inversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text generation models require a large amount of data annotation and modeling resources, making them inefficient in generating text without significant resource consumption.
Innovation Solution
A method and device utilizing a self-encoding neural network that obtains a text word vector, reversely inputs it into a trained model to extract a hidden feature, modifies it based on a classification scale and requirement, and generates text without relying on extensive data annotation and modeling resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If adversarial generation model is used for text generation, then text generation capability is achieved, but data annotation resources and modeling resources are consumed excessively
Solution Approach 1:
The patent inverts the conventional text generation approach by using a self-encoding neural network that encodes text into hidden layer representations and then decodes them back to generate text, rather than using adversarial models that require extensive training data and computational resources. This inversion of the generation paradigm reduces resource consumption while maintaining text generation capability
Solution Approach 2:
The patent changes key parameters of the neural network model by using a self-encoding architecture with specific hidden layer configurations, activation functions, and loss functions that are optimized for text generation. By adjusting these parameters and using a different model architecture, the system achieves text generation with reduced data annotation and modeling resource requirements
Data Source
AI summary
The present disclosure relates to the technical field of natural language understanding, and provides a method, a terminal and a medium for generating a text based on a self-encoding neural network. The method includes: obtaining a text word vector and a classification requirement of a statement to be input; reversely inputting the text word vector into a trained self-encoding neural network model to obtain a hidden feature of an intermediate hidden layer of the self-encoding neural network model; modifying the hidden feature according to a preset classification scale and the classification requirement; defining the modified hidden feature as the intermediate hidden layer of the self-encoding neural network model, and reversely generating a word vector corresponding to an input layer of the self-encoding neural network model by the intermediate hidden layer; and generating the corresponding text, according to the generated word vector.