Summary of Constrained Decoding For Cross-lingual Label Projection, by Duong Minh Le et al.
Constrained Decoding for Cross-lingual Label Projection
by Duong Minh Le, Yang Chen, Alan Ritter, Wei Xu
First submitted to arxiv on: 5 Feb 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper presents a novel approach to improve zero-shot cross-lingual transfer learning for low-resource languages, focusing on fine-grained predictions. Current methods rely on translation and label projection, but state-of-the-art marker-based approaches degrade translation quality due to injected label markers. The authors propose constrained decoding for label projection, which preserves text quality and can be applied to both training and test data strategies. This approach is evaluated on two cross-lingual tasks (Named Entity Recognition and Event Argument Extraction) across 20 languages, outperforming state-of-the-art marker-based methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper tries to solve a problem in machine learning where we want to use models trained on one language to work well on other languages without extra training data. This is hard when we need to make precise predictions about specific words and phrases. The authors suggest a new way to do this by using special rules for decoding labels, which helps keep the translation quality good. They test their idea on two important tasks (finding named entities and extracting event information) across 20 languages and show that it works better than other methods. |
Keywords
* Artificial intelligence * Machine learning * Named entity recognition * Transfer learning * Translation * Zero shot
