Summary Generation With Knapsack-Based Sentence Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automatic summarization technologies using neural networks generate summaries that do not consider factors beyond sentence profitability, resulting in low-quality summaries that fail to meet user reading requirements, thereby reducing user experience.
Innovation Solution
A method that scores sentences based on both profit and length using a knapsack problem approach, selecting sentences with the highest score sum and length sum less than a threshold to form a high-quality summary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If only sentence profit is considered in summary generation, then the summary can be generated quickly using simple neural network scoring, but the summary quality deteriorates and does not meet user reading requirements
Solution Approach 1:
The patent extends the summary generation approach from considering only sentence profit (one dimension) to simultaneously considering both sentence profit and sentence length (two dimensions). This is achieved by formulating the problem as a knapsack optimization where sentences are selected based on both their profit scores and length constraints, thereby resolving the contradiction between generation speed and quality by adding a new optimization dimension without significantly increasing computational complexity
Solution Approach 2:
The patent changes the optimization parameters from single-dimensional profit maximization to multi-dimensional optimization involving both profit and length. By introducing length as a constraint parameter in the knapsack problem formulation, the system achieves higher summary quality while maintaining efficient generation through dynamic programming or greedy algorithms that handle these additional parameters systematically
2Manufacturing precision
If multiple factors including both profit and length are considered in summary generation, then summary quality improves and meets user requirements, but the generation process becomes more complex
Solution Approach 1:
The patent extracts the length consideration from the general summary generation process and incorporates it as a specific constraint parameter in the knapsack problem formulation. By separating the profit scoring function from the length constraint, the system maintains modular architecture where the neural network handles profit assessment while the optimization layer manages length constraints, thereby improving quality without proportionally increasing overall system complexity
Solution Approach 2:
The patent introduces an intermediary optimization layer that mediates between the neural network profit scoring and the final summary selection. This intermediary knapsack optimization module processes both profit scores and length constraints to produce the final sentence selection, acting as a bridge that integrates multiple factors systematically without requiring direct complex interactions between all components
Data Source
AI summary
This application discloses a summary generation method and a related device thereof, to automatically generate a summary that is of a target text with high quality and that meets a length limit. The method in this application includes: obtaining a target text, where the target text includes N sentences, and N≥2; scoring the N sentences by using an extractive summarization model, to obtain scores of the N sentences, where the scores of the N sentences indicate profits of the N sentences in the target text; determining, from the N sentences based on the scores of the N sentences and lengths of the N sentences, M sentences whose score sum is largest and whose length sum is less than a length threshold, where N≥M≥1; and generating a summary of the target text based on the M sentences.