Chatbot Intent Training Example Generation Using Similar Utterances
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The process of gathering training examples for new intents in chatbots is laborious and often lacks sufficient data, making it inefficient and time-consuming.
Innovation Solution
A system and method that utilizes controllable text generation to automatically create training examples for new intents by searching for similar existing intents, generating text based on meta-knowledge, and validating the examples using external documents.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual data gathering is used to create training examples for new intents, then the quality and accuracy of training data can be maintained, but the process becomes laborious and time-consuming
Solution Approach 1:
The system copies existing training examples from similar intents in the database. When creating training examples for a new intent, the system identifies intents with similar meta-knowledge and replicates their training examples, adapting them to the new intent context. This copying approach eliminates manual data gathering while preserving data quality through proven templates.
Solution Approach 2:
The system performs preliminary actions by pre-processing and storing meta-knowledge from existing intents in a structured database format. Before new intents are created, the system has already organized similar examples and their metadata, enabling rapid retrieval and adaptation. This preliminary organization reduces the time required for new intent creation while maintaining accuracy.
2Reliability
If more training examples are gathered manually to improve model performance, then the chatbot's accuracy improves, but the labor and resources required increase significantly
Solution Approach 1:
The system creates universal training examples that can serve multiple intents. By identifying meta-knowledge patterns across similar intents, the system generates training examples with placeholder variables that can be adapted to different specific intents. This multi-functional approach allows one set of training examples to benefit multiple chatbot functions, improving reliability without proportional increases in manual effort.
Solution Approach 2:
The system copies and adapts training examples from existing intents to create new ones, leveraging proven successful patterns. Instead of creating unique training examples for each intent, the system replicates effective examples from similar contexts, ensuring reliable performance while dramatically reducing the productivity burden of manual creation.
3Productivity
If training examples are generated automatically without validation, then the process speed increases, but the quality and reliability of generated examples may decrease
Solution Approach 1:
The system implements feedback mechanisms where generated training examples are validated against the original intent definitions and similar existing intents. The validation process provides feedback on whether generated examples meet quality standards, and unsuccessful examples are regenerated or adjusted. This feedback loop maintains high quality while preserving the speed benefits of automated generation.
Solution Approach 2:
The system performs self-validation of generated training examples by comparing them against stored meta-knowledge and similarity criteria. The automated system serves its own quality control needs without requiring external manual review, maintaining both high speed and quality through self-correcting generation processes that verify their own output against established standards.
Data Source
AI summary
A topic for building a new intent on which to train a chatbot can be received. A database of chatbot training data can be searched for a candidate intent having meta-knowledge similar to the received topic. Utterances associated with the candidate intent can be extracted. The received topic and the extracted utterances can be input to a trained machine learning model. The trained machine learning model generates example utterances for the new intent. The new intent with the generated example utterances can be used as training data for training the chatbot.