Summary of To Label or Not to Label: Hybrid Active Learning For Neural Machine Translation, by Abdul Hameed Azeemi et al.
To Label or Not to Label: Hybrid Active Learning for Neural Machine Translation
by Abdul Hameed Azeemi, Ihsan Ayyub Qazi, Agha Ali Raza
First submitted to arxiv on: 14 Mar 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 proposes a novel active learning strategy called Hybrid Uncertainty and Diversity Sampling (HUDS) for domain adaptation in neural machine translation (NMT). HUDS combines the benefits of uncertainty sampling, which selects instances with high model uncertainty, and diversity sampling, which chooses heterogeneous examples. The approach computes uncertainty scores for unlabeled sentences, stratifies them, clusters sentence embeddings within each stratum, and calculates diversity scores by distance to the centroid. A weighted hybrid score is then used to select the top instances for annotation in each active learning iteration. Experimental results on multi-domain German-English and French-English datasets show that HUDS outperforms other strong active learning baselines. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Active learning (AL) helps reduce labeling costs by selecting small subsets from unlabeled data for annotation. Two popular methods are diversity sampling, which picks varied but trivial examples, and uncertainty sampling, which chooses repetitive, uninformative instances. To overcome these limitations, the paper introduces HUDS, a strategy that combines both approaches. It calculates uncertainty scores, stratifies sentences, clusters embeddings, and computes diversity scores to select the best instances for annotation. The results on German-English and French-English datasets show that HUDS is better than other AL methods. |
Keywords
* Artificial intelligence * Active learning * Domain adaptation * Translation
