Syllable-Based Multi-POSMORPH Annotation for Korean Morphological Analysis and Part-of-Speech Tagging

Various research approaches have attempted to solve the length difference problem between surface form and base of words in Korean morphological analysis part-of-speech (POS) tagging task. The compound POS method is a popular approach, which tackles using annotation tags. However, dictionary required for post-processing recover dissolve ambiguity tags, degrades system performance. In this study, we propose novel syllable-based multi-POSMORPH within one framework, without post-processing. A tag created by combining tags morpheme syllables simultaneous recovery. model implemented with two-layer transformer encoder, lighter than existing models based on large language models. Nonetheless, experiments demonstrate that performance proposed comparable to, or better than, previous