uzpostagger-latin-3

This model is a fine-tuned version of tahrirchi/tahrirchi-bert-base on uzbekpos dataset. It achieves the following results on the evaluation set:

• Loss: 0.3122
• Precision: 0.8858
• Recall: 0.8625
• F1: 0.8740
• Accuracy: 0.9227

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	1.0	25	1.0019	0.5689	0.3641	0.4440	0.6773
No log	2.0	50	0.5215	0.7924	0.7209	0.7549	0.8575
No log	3.0	75	0.3725	0.8629	0.8321	0.8472	0.9073
No log	4.0	100	0.3285	0.8751	0.8458	0.8602	0.9150
No log	5.0	125	0.3122	0.8858	0.8625	0.8740	0.9227

Framework versions

• Transformers 4.32.1
• Pytorch 2.2.0
• Datasets 2.17.1
• Tokenizers 0.13.3

Citation Information

@inproceedings{bobojonova-etal-2025-bbpos,
    title = "{BBPOS}: {BERT}-based Part-of-Speech Tagging for {U}zbek",
    author = "Bobojonova, Latofat  and
      Akhundjanova, Arofat  and
      Ostheimer, Phil Sidney  and
      Fellenz, Sophie",
    editor = "Hettiarachchi, Hansi  and
      Ranasinghe, Tharindu  and
      Rayson, Paul  and
      Mitkov, Ruslan  and
      Gaber, Mohamed  and
      Premasiri, Damith  and
      Tan, Fiona Anting  and
      Uyangodage, Lasitha",
    booktitle = "Proceedings of the First Workshop on Language Models for Low-Resource Languages",
    month = jan,
    year = "2025",
    address = "Abu Dhabi, United Arab Emirates",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.loreslm-1.23/",
    pages = "287--293",
    abstract = "This paper advances NLP research for the low-resource Uzbek language by evaluating two previously untested monolingual Uzbek BERT models on the part-of-speech (POS) tagging task and introducing the first publicly available UPOS-tagged benchmark dataset for Uzbek. Our fine-tuned models achieve 91{\%} average accuracy, outperforming the baseline multi-lingual BERT as well as the rule-based tagger. Notably, these models capture intermediate POS changes through affixes and demonstrate context sensitivity, unlike existing rule-based taggers."
}