Document Summarization Using Submodular Greedy Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sentence compression summarization methods are limited by restricted object functions and incur high calculation costs, making it difficult to efficiently summarize large-scale document data.
Innovation Solution
A document summarizing device and method that utilize submodular maximization with a greedy algorithm to select non-redundant subsets from a dependency structure tree, maximizing information retention while adhering to a length constraint, using a submodular function to evaluate the importance of document subsets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If sentence compression summarization is used to handle long sentences with redundant portions, then a summary with a large amount of information can be created within length constraints, but the types of applicable object functions are limited and the approach of obtaining a better summary by appropriately designing object functions cannot be employed
Solution Approach 1:
The patent changes the parameter of object function design by introducing a configurable objective function that can be adjusted through parameters such as importance weights for different sentences, redundancy thresholds, and information density metrics. This allows the summarization system to adapt to different information retention requirements while maintaining computational efficiency.
2Loss of information
If sentence compression summarization is used to create summaries with large amounts of information, then length constraints are satisfied, but processing that incurs large calculation costs (integer programming, dynamic programming) is required making application to summary creation of large-scale document data difficult
Solution Approach 1:
The patent extracts only the essential computational elements from complex optimization methods, using a simplified greedy algorithm that selects sentences based on pre-calculated importance scores and redundancy metrics. This extraction of core functionality eliminates the need for computationally expensive integer programming or dynamic programming while maintaining effective information retention.
Solution Approach 2:
The patent employs lightweight, computationally inexpensive scoring functions that can be rapidly calculated and discarded for each sentence evaluation. These simple objective functions serve as disposable computational tools that provide sufficient guidance for sentence selection without the overhead of complex optimization algorithms, enabling fast processing of large-scale documents.
3Length of stationary object
If only a few redundant and long sentences are selected using sentence extraction summarization, then the upper limit of length constraint is reached, but a summary with a large amount of information has not been able to be obtained
Solution Approach 1:
The patent segments long sentences with redundant portions into multiple shorter, information-dense sentence units. By dividing the document into finer-grained segments and evaluating each independently using the objective function, the system can select more sentences within the length constraint, thereby increasing the total information content while maintaining the summary within the specified length limit.
Data Source
AI summary
The present invention relates to summarization of a document. The method includes receiving texts of the document and a submodular function for defining an amount of information of a subset of the document as a summary of the document, generating a dependency structure tree representing a dependency structure of sentences in a document, identifying a set of paths including the root of the dependency structure tree, recursively selecting a paths using a greedy algorithm for maximizing the submodular function, and outputs a subset with the maximum amount of information as a the summary of the document.