The present study extends recent work on Universal Dependencies annotations for second-language (L2) Korean by introducing a semi-automated framework that identifies morphosyntactic constructions from XPOS sequences and aligns those constructions with corresponding UPOS categories. We also broaden the existing L2-Korean corpus by annotating 2,998 new sentences from argumentative essays. To evaluate the impact of XPOS-UPOS alignments, we fine-tune L2-Korean morphosyntactic analysis models on datasets both with and without these alignments, using two NLP toolkits. Our results indicate that the aligned dataset not only improves consistency across annotation layers but also enhances morphosyntactic tagging and dependency-parsing accuracy, particularly in cases of limited annotated data.

However, Zhang et al. (2020) found that We present an error analysis of neural the only way to leverage POS tags (both coarse UPOS taggers to evaluate why using gold and fine-grained) for English and Chinese dependency standard tags has such a large positive contribution parsing was to utilise them as an auxiliary to parsing performance while using task in a multi-task framework. Further, Anderson predicted UPOS tags either harms performance and Gómez-Rodríguez (2020) investigated the impact or offers a negligible improvement. UPOS tagging accuracy has on graph-based We evaluate what neural dependency and transition-based parsers and found that a prohibitively parsers implicitly learn about word types high tagging accuracy was needed to and how this relates to the errors taggers utilise predicted UPOS tags. Here we investigate make to explain the minimal impact using whether dependency parsers inherently learn similar predicted tags has on parsers. We also word type information to taggers, and therefore present a short analysis on what contexts can only benefit from the hard to predict tags that result in reductions in tagging performance.