uhhlt/am-roberta

Introduction

This is the Amharic RoBERTa transformer-based LM. It is part of the effort to build benchmark datasets and models for Amharic NLP.

Examples

If you want to test the model in the Hosted inference API, copy the following texts to the box (right side)

Example 1:

አበበ <mask> በላ ።

Example 2:

የአገሪቱ አጠቃላይ የስንዴ አቅርቦት ሶስት አራተኛው የሚመረተው በአገር <mask> ነው።

The example shows possible words for the fill in the blank -- mask task

Resources and publication

More resource regarding Amharic NLP is available here If you use the model in your work, please cite the following paper

@Article{fi13110275,
AUTHOR = {Yimam, Seid Muhie and Ayele, Abinew Ali and Venkatesh, Gopalakrishnan and Gashaw, Ibrahim and Biemann, Chris},
TITLE = {Introducing Various Semantic Models for Amharic: Experimentation and Evaluation with Multiple Tasks and Datasets},
JOURNAL = {Future Internet},
VOLUME = {13},
YEAR = {2021},
NUMBER = {11},
ARTICLE-NUMBER = {275},
URL = {https://www.mdpi.com/1999-5903/13/11/275},
ISSN = {1999-5903},
DOI = {10.3390/fi13110275}
}