Deep Learning Text Correction via Knowledge Base Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep learning-based text generation systems often produce text that is prone to randomness and incorrectness, leading to issues with truthfulness and coherence, especially in open-domain content generation where facts are distorted and meaningless sentences are generated.
Innovation Solution
A computer-implemented method that corrects raw text generated by deep learning techniques by using a search query to retrieve candidate true sentences from a knowledge base or corpus, aligning them syntactically and semantically, and replacing entities or phrases to generate corrected text while maintaining the original structure and coherence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If deep learning techniques are used to generate text, then text generation speed and productivity are improved, but text accuracy and truthfulness deteriorate
Solution Approach 1:
The system implements a feedback mechanism where generated text is automatically verified against knowledge bases and search results. The verification module provides feedback on factual accuracy, and the correction module uses this feedback to generate corrected text, thereby improving truthfulness while maintaining high generation speed through automated processes
Solution Approach 2:
The system performs preliminary actions by pre-creating knowledge bases and search query structures before text generation occurs. This allows the verification and correction processes to run efficiently alongside generation, ensuring accuracy without significant speed penalty
2Adaptability or versatility
If deep learning techniques are used to generate text, then content generation versatility is improved, but sentence coherence and meaning deteriorate
Solution Approach 1:
The verification module provides feedback on sentence coherence and meaningfulness by comparing generated text against knowledge bases. The correction module uses this feedback to rewrite incoherent or meaningless sentences while preserving the versatile content generation capabilities through automated correction processes
Solution Approach 2:
The system copies successful patterns from verified true sentences and knowledge base entries to correct generated text. By copying coherent structures and meaningful content from reliable sources, the system maintains versatility while improving coherence through systematic replication of proven patterns
3Reliability
If correction processes are added to deep learning text generation, then text truthfulness is improved, but processing time and complexity increase
Solution Approach 1:
The system performs preliminary actions by pre-creating knowledge bases and search query structures before text generation. This allows verification and correction to occur during generation rather than as separate post-processing steps, minimizing additional processing time while ensuring truthfulness
Solution Approach 2:
The verification and correction processes operate continuously alongside text generation rather than sequentially. The correction module generates corrected text during the same time frame as the original generation, ensuring continuous useful action and minimizing total processing time while improving truthfulness
Data Source
AI summary
Methods for correcting raw text generated by deep learning techniques is disclosed. The methods may be performed by systems/computing devices described herein. Raw text previously generated by the deep learning techniques may be obtained. A search query can be generated from a raw text sentence of the raw text. The search query is executed against a knowledge base or a corpus of text to obtain a set of search results, the set of search results comprising a plurality of candidate true sentences that can potentially be utilized to correct one or more entities or phrases of the raw text sentence. A candidate true sentence is selected from the plurality and used to correct the raw text sentence. For example, at least one entity or phrase of the candidate true sentence can be used to replace a corresponding entity or phrase of the raw text sentence.