Datasets:
dataset_info:
features:
- name: hindi_token
sequence: string
- name: hindi_upos
sequence:
class_label:
names:
'0': ADJ
'1': ADP
'2': ADV
'3': AUX
'4': CCONJ
'5': DET
'6': INTJ
'7': NOUN
'8': NUM
'9': PART
'10': PRON
'11': PROPN
'12': PUNCT
'13': SCONJ
'14': SYM
'15': VERB
'16': X
- name: angika_token
sequence: string
- name: angika_upos
sequence:
class_label:
names:
'0': ADJ
'1': ADP
'2': ADV
'3': AUX
'4': CCONJ
'5': DET
'6': INTJ
'7': NOUN
'8': NUM
'9': PART
'10': PRON
'11': PROPN
'12': PUNCT
'13': SCONJ
'14': SYM
'15': VERB
'16': X
- name: magahi_token
sequence: string
- name: magahi_upos
sequence:
class_label:
names:
'0': ADJ
'1': ADP
'2': ADV
'3': AUX
'4': CCONJ
'5': DET
'6': INTJ
'7': NOUN
'8': NUM
'9': PART
'10': PRON
'11': PROPN
'12': PUNCT
'13': SCONJ
'14': SYM
'15': VERB
'16': X
- name: bhojpuri_token
sequence: string
- name: bhojpuri_upos
sequence:
class_label:
names:
'0': ADJ
'1': ADP
'2': ADV
'3': AUX
'4': CCONJ
'5': DET
'6': INTJ
'7': NOUN
'8': NUM
'9': PART
'10': PRON
'11': PROPN
'12': PUNCT
'13': SCONJ
'14': SYM
'15': VERB
'16': X
splits:
- name: test
num_bytes: 777906
num_examples: 707
download_size: 162976
dataset_size: 777906
configs:
- config_name: default
data_files:
- split: test
path: data/test-*
task_categories:
- token-classification
language:
- hi
- anp
- bho
- mag
pretty_name: Bihari Languages UPOS Dataset (Angika, Magahi, Bhojpuri)
size_categories:
- n<1K
license: cc-by-nc-sa-4.0
Bihari Languages UPOS Dataset
This dataset provides Part-of-Speech (POS) tags for Angika (anp), Magahi (mag), and Bhojpuri (bho), parallelly aligned with Hindi (hi). The annotations follow the Universal Dependencies (UD) Universal Part-of-Speech (UPOS) standard.
This work is part of research conducted at the Department of Computer Science and Engineering, IIT Bombay.
Dataset Details
- Languages: Angika, Magahi, Bhojpuri, Hindi
- Task: Token Classification (Part-of-Speech Tagging)
- Schema: Universal Dependencies (UPOS)
- Total Tags: 18 (Standard 17 UPOS tags + 1 UNK/X)
Supported Tags
The dataset uses the following integer mapping for the test split:
0: NOUN, 1: PUNCT, 2: ADP, 3: NUM, 4: SYM, 5: SCONJ, 6: ADJ, 7: PART, 8: DET, 9: CCONJ, 10: PROPN, 11: PRON, 12: UNK, 13: X, 14: ADV, 15: INTJ, 16: VERB, 17: AUX
Institutional Credit & Support
- This research was conducted at the Department of Computer Science and Engineering, IIT Bombay.
- The work is supported by a Ph.D. grant from the TCS Research Foundation for research on extremely low-resource Indian languages.
🚀 Getting Started
You can load the dataset directly using the Hugging Face datasets library:
from datasets import load_dataset
# Load the test split
dataset = load_dataset("snjev310/bihari-languages-upos", split="test")
# Access the first sentence in Angika
print(f"Tokens: {dataset[0]['angika_token']}")
print(f"UPOS IDs: {dataset[0]['angika_upos']}")
# Map integer IDs back to tag names
labels = dataset.features["angika_upos"].feature.names
readable_tags = [labels[i] for i in dataset[0]['angika_upos']]
print(f"UPOS Tags: {readable_tags}")
Contact
Sanjeev Kumar CSE IIT Bombay
Email: sanjeev@cse.iitb.ac.in
Research & Citation
If you use this dataset in your research, please cite the following paper published in ACL 2024:
@inproceedings{kumar-etal-2024-part,
title = "Part-of-speech Tagging for Extremely Low-resource {I}ndian Languages",
author = "Kumar, Sanjeev and
Jyothi, Preethi and
Bhattacharyya, Pushpak",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.857/",
doi = "10.18653/v1/2024.findings-acl.857",
pages = "14422--14431"
}