Explainable Text Summarization With Source-Traceable AI Output

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Large Language Models (LLMs) generate factually incorrect information, leading to a lack of trust and reliability, and are computationally inefficient for text summarization tasks, especially in high-risk scenarios.

Innovation Solution

A method that extracts a subset of sentences from an input document, adds context, and uses a generative language model to generate a fluent summary while tracing the source information, providing a transparent summary view.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If a generative language model is used to create text summaries, then the summary can be generated fluently and efficiently, but the model may generate false information (hallucinations) that reduces trust and reliability

Engineering Contradiction:

Improvesummary generation efficiencyVSAvoidfactual accuracy

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The summarization process is divided into two distinct stages: first, an extractive summarization step that selects and extracts only factual sentences from the input document; second, an abstractive summarization step that uses the extracted sentences as ground truth to generate fluent summaries. This segmentation prevents the model from hallucinating by ensuring all generated content is rooted in verified source material.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The extracted sentences from the first stage serve as an intermediary bridge between the input document and the final summary. These extracted sentences act as ground truth constraints that guide the second stage model, ensuring the generated summary remains factually accurate while maintaining fluency and readability.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Loss of information

If the entire input document is processed by the generative language model, then comprehensive information can be captured, but computational resources and time are consumed inefficiently

Engineering Contradiction:

Improveinformation completenessVSAvoidcomputational resource consumption

Core Design Contradiction:

Loss of informationVSUse of energy by moving object

Solution Approach 1:

The system first extracts only the essential information (key sentences) from the input document using extractive summarization. This extraction step filters out redundant and less important content, creating a condensed representation that retains the core information while significantly reducing the amount of data that needs to be processed by the computationally intensive generative model in the second stage.

Inventive Principle:

Principle #2Taking out (Extraction)

Data Source

PatentUS12632659B2Explainable and efficient text summarization

Publication Date: 2026.05.19 NEC CORP

AI summary

A computer-implemented, machine learning method for generating explainable text summaries includes extracting a subset of sentences from an input document as an extractive summary and adding context to the extracted sentences to generate a prompt. A fluent summary is generated by using the prompt as input to a generative language model. Source information for a sentence from the fluent summary is determined by mapping the sentence from the fluent summary to a sentence in the extractive summary and the sentence from the extractive summary to a sentence from the input document. A transparent summary view is generated showing the sentence from the fluent summary along with the source information from the extractive summary and the input document for display on a user interface. The method has applications including, but not limited to medical AI, public safety and other machine learning applications for reliable and explainable document summarization.