LLM Intent Narratives With Confidence Checks for Call Transcripts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining the intent of conversational interactions rely heavily on human-generated summaries, which are often incomplete or inaccurate, leading to inefficiencies and inaccuracies in extracting valuable information from customer support interactions.
Innovation Solution
Utilizing large language models with engineered prompts to detect and generate a fluent, complete narrative of the intent expressed in conversational interactions, followed by rule-based checks and adjustments to ensure accuracy and readability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human-generated summaries are used to determine intent, then the process is simple to implement, but the accuracy and completeness of intent extraction deteriorates
Solution Approach 1:
The patent introduces large language models as intermediary components between the interaction transcript and the intent determination process. The LLM acts as a mediator that transforms raw transcript data into structured intent narratives, thereby improving extraction accuracy while managing system complexity through automated processing.
Solution Approach 2:
The patent replaces the manual mechanical process of human-generated summaries with an automated computational system using large language models. This substitution eliminates human intervention in the summarization process, significantly improving intent extraction accuracy and consistency while reducing manual labor requirements.
2Productivity
If manual summarization by representatives is used, then the system complexity is low, but the productivity and efficiency of information extraction deteriorates
Solution Approach 1:
The system enables self-service by allowing the large language model to automatically generate intent narratives without human intervention. The LLM processes interaction transcripts independently, extracting and summarizing intent information autonomously, which dramatically improves productivity while the modular architecture manages system complexity.
Solution Approach 2:
The patent replaces the manual mechanical summarization process with an automated AI-driven system. This substitution eliminates the time-consuming manual effort of representatives writing summaries, thereby significantly improving information extraction efficiency and productivity across large volumes of interactions.
3Measurement precision
If large language models are used to generate intent narratives, then the accuracy and readability of intent summaries improves, but the processing time and computational resources increase
Solution Approach 1:
The system performs preliminary actions by using engineered prompts that are pre-designed to guide the large language model toward accurate and concise intent narratives. These pre-configured prompts optimize the LLM's processing efficiency, reducing the time required to generate high-quality intent summaries while maintaining accuracy.
Solution Approach 2:
The patent employs parameter changes by adjusting the temperature, max tokens, and other generation parameters of the large language model to optimize the balance between narrative accuracy and processing speed. By carefully tuning these parameters, the system achieves high-quality intent extraction while minimizing computational overhead and processing time.
4Reliability
If multiple large language models are used for intent detection, then the reliability and accuracy of intent determination improves, but the device complexity and computational cost increases
Solution Approach 1:
The patent merges multiple large language models into a unified intent determination system where models work collaboratively. By combining the outputs of multiple LLMs and applying confidence score thresholds, the system achieves higher reliability and accuracy in intent detection while managing complexity through integrated processing and shared infrastructure.
Data Source
AI summary
A method for providing an intent expressed in a conversational interaction in narrative form includes detecting a first intent, with a first large language model, from an input comprising at least an interaction transcript and an engineered prompt; generating, with the first large language model, a first narrative for the first intent and a first confidence score corresponding to a probability that the first intent is an intention in fact expressed in the interaction transcript; determining that the first confidence score is greater than or equal to a threshold; checking that the first narrative conforms to fluency and completeness rules, when the first confidence score is greater than or equal to the threshold; and outputting the first narrative as an output detected intent, when the first confidence score is greater than or equal to the threshold and when the first narrative conforms to the fluency and completeness rules.