Few-Shot and Zero-Shot Learning for Historical Text Normalization
| Autor: | Anders Søgaard, Natalia Korchagina, Marcel Bollmann |
|---|---|
| Přispěvatelé: | University of Zurich |
| Rok vydání: | 2019 |
| Předmět: |
FOS: Computer and information sciences
Computer Science - Machine Learning Computer Science - Computation and Language Training set Computer science business.industry Shot (filmmaking) Lemmatisation 410 Linguistics 000 Computer science knowledge & systems Zero shot learning computer.software_genre Machine Learning (cs.LG) Task (project management) 10105 Institute of Computational Linguistics Identity (object-oriented programming) Text normalization Artificial intelligence Baseline (configuration management) business Computation and Language (cs.CL) computer Natural language processing |
| Zdroj: | DeepLo@EMNLP-IJCNLP Bollmann, M, Korchagina, N & Søgaard, A 2019, Few-Shot and Zero-Shot Learning for Historical Text Normalization . in Proceedings of the 2nd Workshop on Deep Learning Approaches for Low-Resource NLP (DeepLo 2019) . Association for Computational Linguistics, pp. 104-114, 2nd Workshop on Deep Learning Approaches for Low-Resource NLP (DeepLo), Hong Kong, China, 03/11/2019 . https://doi.org/10.18653/v1/D19-6112 |
| DOI: | 10.18653/v1/d19-6112 |
| Popis: | 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. Comment: Accepted at DeepLo-2019 |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|