Intent Prediction Augmentation for Entity Value Robustness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models for chatbots are sensitive to entity value changes, leading to inconsistent predictions and erroneous intent classification due to reliance on entity values as shortcuts for prediction.
Innovation Solution
Implement a combination of positive and negative entity-aware data augmentation techniques, including a two-stage augmentation process, to enhance the robustness of intent prediction models by focusing on overall context rather than individual entity value changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If entity values are used as shortcuts for prediction in intent prediction models, then prediction speed is improved, but prediction consistency and accuracy deteriorate due to sensitivity to entity value changes
Solution Approach 1:
The patent introduces entity-aware data augmentation as an intermediary process between training data preparation and model training. This augmentation technique acts as a mediator that transforms original training data into augmented training data by systematically varying entity values, thereby teaching the model to focus on contextual patterns rather than entity shortcuts. The augmented data serves as a bridge that connects fast entity-based prediction with reliable context-based prediction.
Solution Approach 2:
The patent applies entity-aware data augmentation before model training to pre-process the training data. By performing this augmentation in advance, the model is pre-exposed to varied entity values and contextual patterns, enabling it to learn robust prediction behaviors before actual deployment. This preliminary action ensures the model develops resistance to entity value changes before encountering real-world variations.
2Reliability
If entity-aware data augmentation is applied to improve model robustness, then prediction consistency is improved, but training data processing complexity increases
Solution Approach 1:
The patent segments the data augmentation process into distinct operational stages: entity extraction from training utterances, systematic entity value replacement with alternatives, and generation of augmented training datasets. This segmentation allows each step to be independently optimized and managed, reducing overall processing complexity while maintaining robustness improvements.
Solution Approach 2:
The patent employs parameter changes by systematically varying entity values in training data according to predefined augmentation rules. Instead of complex structural transformations, the method changes specific parameters (entity values) within the existing data framework, simplifying the augmentation process while effectively teaching the model to ignore entity shortcuts and focus on contextual patterns.
Data Source
AI summary
Novel techniques are described for positive entity-aware augmentation using a two-stage augmentation to improve the stability of the model to entity value changes for intent prediction. In one particular aspect, a method is provided that includes accessing a first set of training data for an intent prediction model, the first set of training data comprising utterances and intent labels; applying one or more positive data augmentation techniques to the first set of training data, depending on the tuning requirements for hyper-parameters, to result in a second set of training data, where the positive data augmentation techniques comprise Entity-Aware (“EA”) technique and a two-stage augmentation technique; combining the first set of training data and the second set of training data to generate expanded training data; and training the intent prediction model using the expanded training data.