iSEA: An Interactive Pipeline for Semantic Error Analysis of NLP Models
This addresses the challenge for NLP model developers in diagnosing errors without pre-defined hypotheses, though it is incremental as it builds on existing subpopulation-based approaches.
The paper tackles the problem of error analysis in NLP models by proposing iSEA, an interactive pipeline that automatically discovers semantically-grounded subpopulations with high error rates, enabling model developers to understand, validate, and test hypotheses about errors through a human-in-the-loop system.
Error analysis in NLP models is essential to successful model development and deployment. One common approach for diagnosing errors is to identify subpopulations in the dataset where the model produces the most errors. However, existing approaches typically define subpopulations based on pre-defined features, which requires users to form hypotheses of errors in advance. To complement these approaches, we propose iSEA, an Interactive Pipeline for Semantic Error Analysis in NLP Models, which automatically discovers semantically-grounded subpopulations with high error rates in the context of a human-in-the-loop interactive system. iSEA enables model developers to learn more about their model errors through discovered subpopulations, validate the sources of errors through interactive analysis on the discovered subpopulations, and test hypotheses about model errors by defining custom subpopulations. The tool supports semantic descriptions of error-prone subpopulations at the token and concept level, as well as pre-defined higher-level features. Through use cases and expert interviews, we demonstrate how iSEA can assist error understanding and analysis.