Email Thread Segmentation for Token-Limited AI Summaries

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing email summarization technologies face challenges in efficiently processing multiple email threads, exceeding token size limits, generating inaccurate summaries due to duplication or empty content, and experiencing slow response times, leading to disrupted work and missed important information.

Innovation Solution

A method and apparatus utilizing generative artificial intelligence and prompt engineering to divide email threads, apply multiple AI models in parallel, and use re-summarization prompts to ensure accurate and timely summary generation, removing duplicates and unnecessary content, and optimizing for user-selected options.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If the email body is divided according to token size, then the summarization can be processed, but the email body is divided in the middle causing LLM to arbitrarily interpret or discard the cut body, resulting in missing parts or hallucination

Engineering Contradiction:
Improvetoken sizeVSAvoidsummarization accuracy
Core Design Contradiction:
Quantity of substanceVSReliability

Solution Approach 1:

The patent divides email threads into multiple chunks based on token size limits, processing each chunk separately through parallel LLM calls. This segmentation allows handling of large email volumes while maintaining individual chunk integrity within token limits.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a summary correlation mechanism that acts as an intermediary between divided email chunks. This intermediary collects summaries from multiple chunks and correlates them to ensure consistent interpretation across chunk boundaries, preventing arbitrary interpretation or discarding of cut content.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Quantity of substance

If the large body is divided and calling is performed sequentially, then the token size limit is respected, but timeouts occur frequently and the time waiting for a response from the server becomes longer

Engineering Contradiction:
Improvetoken sizeVSAvoidresponse time
Core Design Contradiction:
Quantity of substanceVSLoss of time

Solution Approach 1:

The patent segments the email body into multiple chunks that fit within token size limits, enabling independent processing of each chunk without exceeding model constraints.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements periodic parallel batch processing of email chunks, sending multiple chunks to the LLM simultaneously in periodic batches rather than sequential processing. This periodic parallel action significantly reduces total processing time and eliminates frequent timeouts.

Inventive Principle:
Principle #19Periodic action

3Device complexity

If duplicate parts of the body and an empty email are sent to the generative artificial intelligence, then the processing is simple, but unnecessary relationships may be derived causing meaningless hallucinations or incorrect summarization

Engineering Contradiction:
Improveprocessing complexityVSAvoidsummarization accuracy
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The patent performs preliminary filtering of email chunks to remove duplicates and empty content before sending to the LLM. This preliminary action prevents the LLM from processing redundant information that could lead to hallucinations or incorrect summarization.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent extracts and removes duplicate and empty email chunks from the processing queue before LLM invocation. This extraction ensures that only meaningful, unique content is processed, maintaining high summarization accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP4632648A1Method, apparatus, and computer program for email summarization
Publication Date: 2025.10.15 SAMSUNG SDS CO LTD
  • EP4632648A1 patent drawingFigure 1
  • EP4632648A1 patent drawingFigure 2
  • EP4632648A1 patent drawingFigure 3

AI summary

A processor-implemented method including generating a summarization prompt for email summarization with email content according to an email summarization request, forming email threads by dividing threads of the email content, and summarizing the email content using artificial intelligence, based on the email threads grouped in units of threads and the summarization prompt.