Ossetic-COT: Designing a morphologically annotated corpus and morphological analyzer for Ossetic

2026-07-08 · Source: cs.CL updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, extended

Summary

The Ossetic-COT project introduces the first morphologically annotated corpus for Iron Ossetic, adhering to the Universal Dependencies v2 schema. This corpus comprises 5454 manually annotated sentences from the Iron Ossetic Corpus of Oral Texts, totaling 74032 tokens. It was developed through a meticulous process including filtering, manual segmentation, and mapping to UD v2 guidelines, with language-specific extensions for substantives, auxiliaries, pronouns, and additional cases like Dir, Rcs, and Num. The corpus facilitated the training of a BERT-based morphological analyzer, which achieved a tag accuracy of 95.60%. The BERT model was pre-trained on 1,060,693 unlabeled sentences from the Ossetic National Corpus and fine-tuned, with the multilingual BERT (mBERT) classifier selected as the baseline. Both the dataset and the model are freely available.

Key takeaway

For NLP Engineers developing tools for low-resource languages like Ossetic, this new morphologically annotated corpus and BERT-based analyzer offer a critical foundation. You should integrate the freely available Ossetic-COT dataset and the provided baseline model into your projects. This resource, with its 95.60% tag accuracy, significantly reduces the barrier to developing high-quality Ossetic NLP applications, enabling more precise linguistic analysis and model training.

Key insights

A UD-compliant, morphologically annotated corpus for a low-resource language enables high-accuracy NLP tools.

Principles

• Manual annotation ensures high-quality linguistic data.
• Language-specific extensions are crucial for UD compliance.
• BERT-based models are effective for morphological analysis.

Method

The process involves filtering, manual segmentation, mapping to UD v2 with language-specific extensions, manual refinement, and automated CoNLL-U conversion, followed by BERT pre-training and fine-tuning.

In practice

• Use Ossetic-COT for training neural taggers.
• Apply the BERT analyzer for Ossetic NLP tasks.
• Consult annotation guidelines for complex morphology.

Topics

• Ossetic Language
• Morphological Analysis
• Universal Dependencies
• BERT Models
• Corpus Annotation
• Low-Resource NLP

Code references

• ania3000/Ossetic-COT

Best for: AI Scientist, NLP Engineer, Research Scientist

Related on AIssential

• Curation of a Palaeohispanic Dataset for Machine Learning — A new structured dataset enables machine learning analysis of under-resourced Palaeohispanic languages.
• dialect2vec: A vector-based method for dialectal transcription of Portuguese from ALiB questionnaires — dialect2vec uses byte-level encoding of IPA sequences to model dialectal variation, outperforming subword models.
• A Scalable Tool for Measuring Manner and Result Verbs in Developmental Language Research — A RoBERTa classifier, trained on LLM-generated annotations, accurately distinguishes manner and result verbs at scale.
• Annotation Challenges in Low-Resource African Languages — High-quality NLP for African languages demands culturally-attuned annotation, robust QA, and fair, community-centric practices.
• Morpheus: A Morphology-Aware Neural Tokenizer and Word Embedder for Turkish — Morpheus offers a reversible, morphology-aware neural tokenizer and word embedder specifically for agglutinative languages like Turkish.
• PACUTE: Phonology-, Affix-, and Character-level Understanding of Tokens for Filipino — Subword tokenization hinders LLM understanding of complex morphology, especially in languages like Filipino.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CL updates on arXiv.org.