What type of study is this?

September 10, 2025Big Data and Cognitive ComputingOpen Access

Parsing Old English with Universal Dependencies—The Impacts of Model Architectures and Dataset Sizes

View Full Paper

Ask AI

Bookmark

The pretrained tok2vec model achieves 83.24% UAS and 74.23% LAS, outperforming others significantly.
Medium-complexity architectures show better cost-benefit trade-offs compared to complex transformer models.
Parsing performance peaks at 85–90% accuracy for tagging tasks, while dependency parsing reaches around 75% accuracy.
Insights into scaling laws reveal diminishing returns in performance improvements beyond 10,000 words of data.

Discussion

No takes yet. Share an insight, caveat, or question.

Implication

Empirical comparison of neural architectures shows pretrained models outperform transformers in Old English, suggesting implications for historical language processing.

Cite This Study

Arista et al. (2025) studied this question.

synapsesocial.com/papers/68c19f9c54b1d3bfb60db4f6 https://doi.org/10.3390/bdcc9080199

View Full Paper

Ask AI

Bookmark

Implication

Empirical comparison of neural architectures shows pretrained models outperform transformers in Old English, suggesting implications for historical language processing.

Parsing Old English with Universal Dependencies—The Impacts of Model Architectures and Dataset Sizes

Key Points

Discussion

Implication

Cite This Study

Parsing Old English with Universal Dependencies—The Impacts of Model Architectures and Dataset Sizes

Key Points

Discussion

Implication

Cite This Study

Also Consider

Also Consider