Email Thread Segmentation for Faster AI Summarization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing email summarization technologies face challenges such as difficulty in processing multiple email threads quickly, exceeding token limits of large language models, generating incorrect summaries due to duplicate or empty emails, and prolonged response times when using generative artificial intelligence.
Innovation Solution
A method and apparatus utilizing generative artificial intelligence and prompt engineering to divide email threads, remove duplicates and empty content, and perform parallel calls to generate optimized summaries based on user-selected options, ensuring accurate and efficient summarization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If email threads are divided and processed sequentially, then each thread can be summarized individually, but the total response time becomes excessively long
Solution Approach 1:
The patent segments email threads into manageable chunks that fit within LLM token limits, then processes these segments in parallel across multiple AI models. This segmentation enables both accurate individual thread summarization and efficient parallel processing, resolving the contradiction between summary accuracy and response time.
Solution Approach 2:
The patent transitions from sequential processing (one-dimensional time) to parallel processing across multiple AI models (multi-dimensional concurrent execution). This dimensional change in processing architecture allows simultaneous summarization of multiple email threads, dramatically reducing total response time while maintaining accuracy through individual model processing of each segment.
2Quantity of substance
If the email body exceeds the allowable token size of a large language model, then the email cannot be processed, but cutting the email body arbitrarily may cause missing parts or hallucination
Solution Approach 1:
The patent divides the email body into multiple segments based on thread boundaries, ensuring each segment fits within LLM token limits. By segmenting at meaningful boundaries rather than arbitrarily, the patent maintains summary accuracy while processing large email volumes. Each segment is processed independently by AI models, and results are combined to produce the final summary.
3Productivity
If duplicate parts or empty emails are sent to generative artificial intelligence, then processing continues, but unnecessary relationships are derived and incorrect summaries are generated
Solution Approach 1:
The patent performs preliminary filtering to remove duplicate and empty email threads before they are sent to AI models for processing. This preliminary action ensures that only clean, meaningful data is processed, preventing incorrect summaries from being generated while maintaining high processing throughput by avoiding unnecessary AI calls on problematic data.
4Measurement precision
If multiple AI models are called sequentially, then each model can process the email independently, but the total processing time exceeds the allowable timeout
Solution Approach 1:
The patent implements parallel calling of multiple AI models simultaneously rather than sequentially. This parallel execution across multiple models reduces total processing time dramatically while maintaining summary quality through the collaborative processing of multiple independent models that work on different segments concurrently.
Data Source
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.