Recursive Dialog Summarization for Coherent Long Transcript Summaries

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing abstractive summarization techniques fail to generate coherent and contextually grounded summaries for long dialogs, resulting in disjointed and irrelevant summaries.

Innovation Solution

A machine learning recursive abstractive model is trained to generate summaries by considering each preceding sentence when creating the next sentence, leveraging a divide and conquer approach to improve coherence and relevance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If conventional summarization techniques are used to summarize long dialogs by individually summarizing portions, then processing complexity is reduced, but summary coherence and contextual grounding deteriorate

Engineering Contradiction:

Improveprocessing complexityVSAvoidsummary coherence

Core Design Contradiction:

Device complexityVSReliability

Solution Approach 1:

The patent divides the long dialog into multiple portions or segments that can be processed individually. Each portion is summarized separately, and then these partial summaries are combined to form the complete summary. This segmentation approach reduces processing complexity while maintaining summary coherence through the combination step.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent employs a nested structure where the complete summarization process contains within it multiple levels of partial summarization. Each portion summary is nested within the overall summary framework, allowing the model to build the final coherent summary by integrating these nested partial summaries in a hierarchical manner.

Inventive Principle:

Principle #7Nested doll (Nesting)

2Loss of time

If conventional summarization techniques are used to individually summarize dialog portions, then processing time is reduced, but summary accuracy and relevance deteriorate

Engineering Contradiction:

Improveprocessing timeVSAvoidsummary accuracy

Core Design Contradiction:

Loss of timeVSMeasurement precision

Solution Approach 1:

The dialog is segmented into manageable portions that can be processed in parallel or sequence, reducing overall processing time. The segmentation allows the system to handle long dialogs efficiently while maintaining accuracy through the integration of partial summaries into a coherent whole.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent maintains continuity by ensuring that each partial summary generation process is connected to the overall summarization goal. The model continuously builds upon previous portion summaries, maintaining contextual awareness and ensuring that the final summary accurately reflects the entire dialog rather than losing information through discrete processing.

Inventive Principle:

Principle #20Continuity of useful action

Data Source

PatentUS12547822B2Machine learning divide and conquer techniques for long dialog summarization

Publication Date: 2026.02.10 OPTUM INC

AI summary

Various embodiments of the present disclosure provide summarization techniques for summarizing complex documents, such as long unstructured call transcripts. The summarization techniques include generating a plurality of interaction topics for an interaction transcript and iteratively summarizing each interaction topic based on a preceding partial summary for the interaction transcript that corresponds to a preceding interaction topic that precedes the interaction topic in the interaction transcript. An abstractive summary is generated using a recursive abstractive model that is trained using training data generated based on holistic similarity scores between interaction topics of a call transcript and summary sentences of a corresponding target summary.