We consider Large-Scale Multi-Label Text Classification (LMTC) in the legal domain. We release a new dataset of 57k legislative documents from EURLEX, annotated with ~4.3k EUROVOC labels, which is suitable for LMTC, few- and zero-shot learning. Experimenting with several neural classifiers, we show that BIGRUs with label-wise attention perform better than other current state of the art methods. Domain-specific WORD2VEC and context-sensitive ELMO embeddings further improve performance. We also find that considering only particular zones of the documents is sufficient. This allows us to bypass BERT's maximum text length limit and fine-tune BERT, obtaining the best results in all but zero-shot learning cases.
Large-Scale Multi-Label Text Classification on EU Legislation
In the legal domain, BIGRUs with label-wise attention and BERT fine-tuning using specialized embeddings achieve top results in large-scale multi-label text classification, surpassing state-of-the-art methods.
- Year
- 2019
- Venue
- large-scale-multi-label-text-classification-2
- Authors
- 4
- Hosting
- Abstract onlyARXIV-DEFAULT
Cite
Notes
Only stored in your browser.
Attribution
- Abstract & full text
- arxiv.org/abs/1906.02192ARXIV-DEFAULT
- TL;DR
- Semantic Scholar