LLM Multi-Tool Prompt Correction for Transcription Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional speech recognition systems rely on manual correction for transcription errors, which are inaccurate and not scalable, affecting the performance of downstream computer systems.
Innovation Solution
Employing a large language model (LLM) with a multi-tool prompt to automate the correction process, leveraging machine learning models and search engines to verify and correct transcription errors, and utilizing a memory to store and improve upon previous corrections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual correction is used for transcription errors, then the system can correct errors, but the process is not scalable and affects downstream system performance
Solution Approach 1:
The patent replaces manual mechanical correction processes with an automated LLM-based system. The LLM receives transcription output, analyzes it against the prompt guidelines, and generates corrected transcriptions automatically, eliminating the need for manual intervention while maintaining high accuracy and enabling scalability.
Solution Approach 2:
The system enables self-service correction where the LLM autonomously performs the correction task without human involvement. The multi-tool prompt structure allows the LLM to independently navigate complex correction scenarios, making the system self-sufficient and scalable for large volumes of transcriptions.
2Device complexity
If conventional correction methods are used, then the system structure remains simple, but accuracy and scalability are compromised
Solution Approach 1:
The patent segments the correction process into distinct functional components through the multi-tool prompt structure: error detection, error correction, and quality verification. This segmentation allows each component to be optimized independently while working together to achieve high accuracy, justifying the increased structural complexity.
Solution Approach 2:
The LLM serves multiple functions within the correction system: it detects errors, generates corrections, verifies accuracy, and handles edge cases. This multi-functionality consolidates what would otherwise require multiple separate systems into a single versatile component, improving accuracy while managing complexity through integration.
3Productivity
If automated LLM-based correction is implemented, then scalability and accuracy improve, but computational resources and latency increase
Solution Approach 1:
The system applies partial correction by using the LLM only when and where needed, rather than processing all transcriptions uniformly. The multi-tool prompt enables selective engagement of correction mechanisms based on detected error patterns, reducing unnecessary computational resources while maintaining high scalability for actual error cases.
Solution Approach 2:
The system performs preliminary analysis of the transcription against the prompt guidelines before full correction processing. This preliminary step identifies potential errors and determines the extent of LLM intervention needed, optimizing resource consumption by avoiding unnecessary computational heavy lifting for straightforward transcriptions.
Data Source
AI summary
Techniques for large language model (LLM)-based correction based on a multi-tool prompt are described. In an example, a computer system receives, via a user interface, user input including user-provided information and indicating a request for a task to be performed on the user-provided information. The computer system generates, by using an LLM associated with a prompt, a first input to a first tool based on the user input. The prompt indicating a sequence of steps to perform for the task and tools available to the LLM. The first tool corresponds to a first step of the sequence of steps. The computer system determines, by using the LLM, a first output of the first tool in response to the first input and an update to the user-provided information based on the first output and a completion of the task. The computer system causes the user interface to present the update.