metadata
language:
- hu
tags:
- text-classification
license: apache-2.0
metrics:
- accuracy
widget:
- text: Jó reggelt! majd küldöm az élményhozókat :).
Hungarian Sentence-level Sentiment Analysis Model with huBERT
For further models, scripts and details, see our repository or our demo site.
- Pretrained model used: huBERT
- Finetuned on Hungarian Twitter Sentiment (HTS) Corpus
- Labels: 0 (negative), 1 (positive)
Limitations
- max_seq_length = 128
Results
Model | HTS2 | HTS5 |
---|---|---|
huBERT | 85.56 | 68.99 |
XLM-RoBERTa | 85.56 | 66.50 |
Citation
If you use this model, please cite the following paper:
@inproceedings {yang-bart,
title = {Improving Performance of Sentence-level Sentiment Analysis with Data Augmentation Methods},
booktitle = {Proceedings of 12th IEEE International Conference on Cognitive Infocommunications (CogInfoCom 2021)},
year = {2021},
publisher = {IEEE},
address = {Online},
author = {Laki, László and Yang, Zijian Győző}
pages = {417--422}
}