Data Augmentation for Intent Classification
This work tackles the problem of data scarcity for developers building intent classifiers, but it is incremental as it evaluates existing methods rather than introducing new ones.
The study investigated data augmentation techniques for intent classification to address the high cost of labeled data, finding that some methods significantly improve performance while others have minimal or negative effects.
Training accurate intent classifiers requires labeled data, which can be costly to obtain. Data augmentation methods may ameliorate this issue, but the quality of the generated data varies significantly across techniques. We study the process of systematically producing pseudo-labeled data given a small seed set using a wide variety of data augmentation techniques, including mixing methods together. We find that while certain methods dramatically improve qualitative and quantitative performance, other methods have minimal or even negative impact. We also analyze key considerations when implementing data augmentation methods in production.