Unified Text Segmentation and Summarization Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Long documents lack effective structure, making it difficult for readers to find salient information, especially in the absence of segmentation, which is exacerbated in transcripts of audio/video recordings, hindering extractive summarization.
Innovation Solution
A method that performs section segmentation and summarization simultaneously, using a model that learns robust sentence representations and incorporates an optimization-based regularizer to select diverse and salient sentences, enhanced by the determinantal point process to ensure a comprehensive summary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If long documents are processed without section segmentation, then the summarization process is simpler, but the ability to find and extract salient information deteriorates
Solution Approach 1:
The patent applies segmentation by dividing long documents into sections using predicted section boundaries. The model segments documents into meaningful units (sections) that preserve semantic coherence, enabling readers to navigate and find salient information more effectively while maintaining the integrity of the original content structure.
2Measurement precision
If section segmentation is performed separately from summarization, then each task can be optimized independently, but the overall system complexity increases
Solution Approach 1:
The patent merges section segmentation and summarization into a single unified model that performs both tasks simultaneously. This joint modeling approach reduces system complexity by eliminating the need for separate independent models while maintaining or improving the precision of both segmentation and summarization through shared representations and coordinated optimization.
3Reliability
If multiple models are developed for different tasks (segmentation and summarization), then task-specific performance is improved, but the model development time and computational resources increase
Solution Approach 1:
The patent creates a universal model that performs both section segmentation and summarization functions. This multi-functional model reduces model development time and computational resources by using shared parameters and representations for both tasks, while maintaining task-specific performance through multi-task learning and joint optimization.
4Productivity
If abstractive summarization is used to produce condensed summaries, then summary length is reduced, but factual accuracy and reliability deteriorate due to hallucinations
Solution Approach 1:
The patent applies extraction by selecting and copying salient sentences directly from the source document to form the summary, rather than generating new text. This extractive approach maintains factual accuracy and reliability by preserving the original wording and information, while still achieving summary condensation by selecting only the most important sentences.
Data Source
AI summary
A method including receiving an input comprising natural language texts; segmenting the natural language texts into sections; summarizing the natural language texts; developing a first model based on the plurality of sections and the summary of the natural language texts; identifying one or more salient sentences within the natural language texts using the first model; determining a sentence quality score based on how informative a salient sentence is; determining a sentence similarity score based on a salient sentence's similarity to another salient sentence; developing a second model based on the sentence quality score and the sentence similarity score; combining the first model and the second model into a final model; selecting sentences based on the final model; and generating an extractive summarization using the selected sentences.