Automated NLU Training Data Generation via Logic Model Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing natural-language understanding (NLU) models face challenges in generating effective training data due to the labor-intensive process of manual labeling and the inflexibility of manually generated logic models, which limits their ability to adapt to new language inputs.
Innovation Solution
A computer-implemented method and system for dynamically generating training data by receiving transcripts of customer-agent conversations, selecting responses from a logic model based on agent inputs, determining similarity scores, and training machine learning models only when similarity thresholds are met, thereby automating the labeling process and improving logic model adaptability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual labeling is used to generate training data, then training data can be created, but the process is labor-intensive and time-consuming
Solution Approach 1:
The system uses the logic model itself to automatically generate training data by selecting responses from the logic model and comparing them with agent inputs, eliminating the need for manual labeling. The logic model serves itself by automatically creating training datasets from conversation transcripts.
Solution Approach 2:
The patent replaces the mechanical manual labeling process with an automated computational system that uses similarity scoring algorithms to select and validate training data, substituting human labor with automated machine processes.
2Adaptability or versatility
If manually generated logic models are used, then the model structure is fixed, but the ability to adapt to new language inputs is limited
Solution Approach 1:
The logic model is transformed from a static manually-generated structure to a dynamic system that automatically updates and adapts. The model continuously learns from new conversation transcripts and automatically adjusts its response logic to handle new language inputs effectively.
Solution Approach 2:
The system implements feedback mechanisms where the performance of the logic model is continuously evaluated against actual agent inputs and conversation outcomes, allowing the model to learn from its performance and automatically improve its adaptability to new language patterns.
3Productivity
If automated response selection is implemented, then training data generation is accelerated, but similarity scoring and threshold evaluation add computational complexity
Solution Approach 1:
The system uses similarity scoring as a parameter to automatically evaluate and select responses from the logic model. By transforming the selection process into a quantitative parameter-based comparison, the system achieves automated high-throughput training data generation while managing computational complexity through efficient scoring algorithms.
Data Source
AI summary
A method includes obtaining a transcript of a chat between a customer and an agent, the transcript comprising a customer input from the customer and an agent input from the agent, and selecting, based on the agent input, a response from a plurality of responses representing respective potential replies to the customer input. The method also includes determining that a similarity score between the agent input and the selected response satisfies a similarity threshold, and, based on determining that the similarity score between the agent input and the selected response satisfies the similarity threshold, using the customer input and the selected response to train a machine learning model.