metadata

language:
  - asm
  - bgc
  - bht
  - ckb
  - eng
  - ewe
  - fra
  - guj
  - ibo
  - kan
  - lin
  - luo
  - mal
  - mar
  - nag
  - nde
  - nlx
  - pan
  - peg
  - rus
  - tam
  - tel
  - twi
  - ukr
  - urd
  - vie
  - yor
task_categories:
  - text-classification
  - audio-classification
size_categories:
  - 10K<n<100K

SpeechTaxi

Usage

# pip install datasets pandas soundfile
from datasets import load_dataset
dataset = load_dataset(
    "LennartKeller/SpeechTaxi",
    name="ukr",
    split="train",
    trust_remote_code=True
)

Overview

	Language	alpha3	train	test	dev	total
0	Vietnamese	vie	856	111	106	1073
1	French	fra	851	108	106	1065
2	Russian	rus	822	107	102	1031
3	Ukrainian	ukr	751	97	89	937
4	Kannada	kan	740	100	89	929
5	Gujarati	guj	740	100	89	929
6	Yoruba	yor	739	100	88	927
7	Punjabi	pan	739	100	88	927
8	Naga Pidgin	nag	739	100	89	928
9	Luo (Kenya and Tanzania)	luo	738	100	88	926
10	Tamil	tam	733	100	89	922
11	Marathi	mar	733	99	87	919
12	Assamese	asm	732	98	88	918
13	Haryanvi	bgc	729	100	87	916
14	Bhattiyali	bht	726	98	88	912
15	Malayalam	mal	724	100	89	913
16	Ewe	ewe	724	98	86	908
17	Central Kurdish	ckb	723	93	82	898
18	Telugu	tel	722	96	85	903
19	Igbo	ibo	720	96	87	903
20	Pengo	peg	707	94	86	887
21	Ndebele	nde	699	88	85	872
22	Asante Twi	tw-asante	693	92	88	873
23	Akuapem Twi	tw-akuapem	692	91	84	867
24	Urdu	urd	674	95	80	849
25	Nahali	nlx	672	92	85	849
26	English	eng	569	81	74	724
27	Lingala	lin	560	75	61	696

Citation

@misc{keller2024speechtaximultilingualsemanticspeech,
      title={SpeechTaxi: On Multilingual Semantic Speech Classification}, 
      author={Lennart Keller and Goran Glavaš},
      year={2024},
      eprint={2409.06372},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2409.06372}, 
}