Researchers implemented a neural annotation pipeline for Irabu Ryukyuan to automate the creation of interlinearized text, addressing the high cost of manual documentation. The study evaluated morpheme segmentation, POS tagging, and glossing using small BiLSTM-CRF models under a strict constraint of approximately one hour of annotated discourse.
- Gold POS tags improved grammatical glossing accuracy by 4.4 points on average across five seeds.
- The benefit of POS tags increased as training data decreased, yielding an 11.6-point gain at quarter data volume.
- A POS tier reduced the amount of glossed data required to reach a specific accuracy threshold by more than half.
- In fully automatic pipelines, tagger errors currently prevent realizing these gains, though recovery is projected at higher tagger accuracies.
The authors recommend annotating quadrilinearly (text, POS, gloss, translation) to maximize the efficiency of automated documentation workflows for endangered languages.