Proper names are typically out-of-vocabulary due to their diversity. This is potentially a problem but when Jelínek (2015) simpilfies proper names (along with other 79 groups of words), they've got an improvement of only less than 1% (UAS: 86.12% --> 86.44%, LAS: 79.86% --> 80.58%).
- Cross, J., & Huang, L. (2016). Span-based constituency parsing with a structure-label system and provably optimal dynamic oracles. EMNLP 2016 - Conference on Empirical Methods in Natural Language Processing, Proceedings, (Section 2), 1–11. https://doi.org/10.18653/v1/d16-1001
- Kitaev, N., & Klein, D. (2020). Tetra-Tagging: Word-Synchronous Parsing with Linear-Time Inference, 6255–6261. https://doi.org/10.18653/v1/2020.acl-main.557
- Jelínek, Tomás. "Improvements to Dependency Parsing Using Automatic Simplification of Data." In LREC, vol. 14, pp. 73-77. 2014.