Few-Shot and Zero-Shot Learning for Historical Text Normalization
This work addresses the challenge of data scarcity in historical text normalization for researchers and practitioners, though it is incremental as it systematically compares existing multi-task architectures rather than introducing a new method.
The paper tackled the problem of historical text normalization with limited training data by evaluating 63 multi-task learning configurations across ten datasets from eight languages, showing consistent significant improvements when data is scarce but minimal gains when abundant, and demonstrating that zero-shot learning outperforms a strong identity baseline.
Historical text normalization often relies on small training datasets. Recent work has shown that multi-task learning can lead to significant improvements by exploiting synergies with related datasets, but there has been no systematic study of different multi-task learning architectures. This paper evaluates 63~multi-task learning configurations for sequence-to-sequence-based historical text normalization across ten datasets from eight languages, using autoencoding, grapheme-to-phoneme mapping, and lemmatization as auxiliary tasks. We observe consistent, significant improvements across languages when training data for the target task is limited, but minimal or no improvements when training data is abundant. We also show that zero-shot learning outperforms the simple, but relatively strong, identity baseline.