infoVerse: A Universal Framework for Dataset Characterization with Multidimensional Meta-information
This addresses the need for better dataset analysis in NLP to improve efficiency in tasks like data pruning and active learning, though it is incremental as it builds on existing model-driven meta-information methods.
The paper tackles the problem of dataset characterization by introducing infoVerse, a universal framework that captures multidimensional meta-information to identify valuable samples, resulting in consistent outperformance of baselines in data pruning, active learning, and data annotation applications.
The success of NLP systems often relies on the availability of large, high-quality datasets. However, not all samples in these datasets are equally valuable for learning, as some may be redundant or noisy. Several methods for characterizing datasets based on model-driven meta-information (e.g., model's confidence) have been developed, but the relationship and complementary effects of these methods have received less attention. In this paper, we introduce infoVerse, a universal framework for dataset characterization, which provides a new feature space that effectively captures multidimensional characteristics of datasets by incorporating various model-driven meta-information. infoVerse reveals distinctive regions of the dataset that are not apparent in the original semantic space, hence guiding users (or models) in identifying which samples to focus on for exploration, assessment, or annotation. Additionally, we propose a novel sampling method on infoVerse to select a set of data points that maximizes informativeness. In three real-world applications (data pruning, active learning, and data annotation), the samples chosen on infoVerse space consistently outperform strong baselines in all applications. Our code and demo are publicly available.