Pseudo Labeling Named Entity Recognition Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional chatbots require extensive manual labeling and training data to respond accurately to a wide variety of queries, making them inefficient and labor-intensive, especially when faced with queries not covered in their training data.
Innovation Solution
The system employs pseudo labeling by trained named entity recognition models to automatically generate labeled training data, allowing untrained models to learn and adapt without the need for extensive pre-labeled data, using two or more trained models to apply pseudo labels to user inputs and store matching labels for training, while non-matching inputs are reviewed and corrected.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional chatbots are trained with extensive manual labeled data, then accuracy in responding to queries is improved, but the time and labor required for training increases significantly
Solution Approach 1:
The system enables the chatbot to automatically generate its own training data by having it converse with itself and label the interactions. The chatbot acts as both the service provider and the client, self-generating labeled training data without human intervention, thus resolving the contradiction between accuracy and training time
Solution Approach 2:
The system changes the parameter of data labeling from manual human process to automated AI process. By transforming how training data is labeled (from human-annotated to AI-generated), the system maintains accuracy while dramatically reducing training time and labor requirements
2Adaptability or versatility
If conventional chatbots are trained with extensive manual labeled data, then the variety of queries they can handle is improved, but the labor and resources required for data preparation increases
Solution Approach 1:
The chatbot automatically generates diverse training queries through self-conversation, eliminating the need for manual data collection across multiple domains. This self-service approach expands query variety while reducing the complexity of data preparation
Solution Approach 2:
The system performs preliminary action by having the chatbot generate and label training data before deployment. This advance preparation of diverse queries through automated means reduces the complexity of subsequent data preparation efforts
3Reliability
If manual labeling is used to train chatbots, then coverage of potential cases is improved, but the manpower consumption increases
Solution Approach 1:
The chatbot performs self-labeling of training data, eliminating the need for human annotators. This maintains comprehensive case coverage while dramatically improving training efficiency by removing the manpower bottleneck
Solution Approach 2:
The system replaces the mechanical process of manual labeling with an automated AI-based labeling system. This substitution maintains case coverage while improving productivity by eliminating human labor requirements
4Adaptability or versatility
If chatbots are designed to handle a wide variety of questions, then adaptability is improved, but the system becomes less reliable for specific queries outside the training set
Solution Approach 1:
The system implements continuous self-training where the chatbot constantly generates new training data from its conversations. This continuous improvement cycle allows the system to maintain both wide query range and high response accuracy by continuously adapting to new query types
Solution Approach 2:
The chatbot's training data set is dynamic rather than static, continuously expanding through self-generated conversations. This dynamic approach allows the system to adapt to new query ranges while maintaining reliability through ongoing learning
Data Source
AI summary
Examples of the present disclosure can comprise systems and methods for creating and modifying named entity recognition models. The system can use two or more existing named entity recognition models to output responses to natural language queries for which the models have not yet been trained. When the output from the two or more models match, the query and the resulting output can be stored as training data for a new named entity recognition model. If the output from the two or models do not match, the query and the outputs can be stored in an exceptions file for additional review. In some embodiments, the system can comprise one or more processors and a display for providing a user interface (UI).