Chatbot Neural Network Self-Training via Voice Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing chatbot training methods require significant administrator effort and time to identify and correct misrecognitions, leading to inefficiencies in training and maintenance.
Innovation Solution
An electronic device that automatically trains an artificial intelligence model for a chatbot using voice data from conversations between customers and counselors, allowing for automated identification and correction of misrecognitions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If administrators manually identify and extract learning data from chatbot misrecognitions, then the chatbot can be updated and improved, but a very large amount of administrator effort and time is needed
Solution Approach 1:
The system enables automatic self-training of the chatbot by having the server autonomously generate learning data from conversation recordings, process them through neural networks, and update the chatbot model without requiring administrator intervention. The server automatically identifies customer inquiries, generates appropriate response data, and performs the complete training pipeline, allowing the chatbot to serve its own improvement needs.
Solution Approach 2:
The server acts as an intermediary between the conversation recordings and the chatbot model. It processes the raw conversation data, generates structured learning pairs (inquiry-response), and feeds them to the neural network for training. This intermediary layer automates the data preparation and model update process that would otherwise require manual administrator work.
2Reliability
If administrators manually identify and extract learning data from chatbot misrecognitions, then the chatbot can be updated and improved, but a very large amount of administrator effort is needed
Solution Approach 1:
The system enables automatic self-training of the chatbot by having the server autonomously generate learning data from conversation recordings, process them through neural networks, and update the chatbot model without requiring administrator intervention. The server automatically identifies customer inquiries, generates appropriate response data, and performs the complete training pipeline, allowing the chatbot to serve its own improvement needs.
Solution Approach 2:
The server acts as an intermediary between the conversation recordings and the chatbot model. It processes the raw conversation data, generates structured learning pairs (inquiry-response), and feeds them to the neural network for training. This intermediary layer automates the data preparation and model update process that would otherwise require manual administrator work.
3Measurement precision
If the chatbot is trained with more accurate learning data, then the chatbot's understanding of customer inquiries improves, but the complexity of the training process increases
Solution Approach 1:
Instead of manually creating learning data from misrecognitions, the system inverts the approach by automatically generating learning data from all conversation recordings using neural network classification. The server processes conversations through the chatbot model, identifies misclassified inquiries, and uses these to generate training data, flipping the traditional manual extraction process into an automated computational one.
Solution Approach 2:
The system changes the parameters of the training process by using confidence thresholds and probability values from neural network outputs to automatically identify learning data candidates. By adjusting these parameter thresholds, the system can control the quantity and quality of generated learning data without manual intervention, balancing accuracy with automated processing.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An electronic device is provided. The electronic device includes a memory storing recording data including a content of a conversation and at least one instruction, and a processor configured, by executing the at least one instruction, to input first data corresponding to a first voice in the content of the conversation into a first neural network model and acquire category information of the first data, and acquire category information of second data corresponding to a second voice in the content of the conversation. The processor is configured to, based on the category information of the first data and the category information of the second data being different, train the first neural network model based on the category information of the second data and the first data.