kadirnar/arabic-arena-ar

Arabic Arena: STT benchmark sets (ar)

These are the evaluation sets of Arabic Arena, a multi-dataset WER/CER benchmark for speech-to-text with simulated telephony and call-center conditions.

Each config is a deterministic subset of a public test set: at most 1,000 utterances of 1–30 s, chosen by hash rank. Degraded configs (-noisy, -telephony, -callcenter) are generated with seeded, recorded parameters; see meta.degradation in every row and METHODOLOGY.md.

Columns: id, audio, sampling_rate, text (raw reference), duration, meta (JSON).

from datasets import load_dataset
ds = load_dataset("kadirnar/arabic-arena-ar", "sada-padded", split="test")

Config	Dialect	Domain	Condition	Utterances	Hours	Licence	Source
fleurs_ar-clean	MSA	read	original audio (16 kHz)	428	1.30	CC-BY-4.0	google/fleurs
cv18_ar-clean	MSA	read (crowd-sourced)	original audio (16 kHz)	1000	1.21	CC0-1.0	MohamedRashad/common-voice-18-arabic
masc_clean-clean	mixed	youtube	original audio (16 kHz)	1000	1.18	CC-BY-4.0	MohamedRashad/MASC-Arabic
sada-clean	Gulf (Saudi)	tv shows	original audio (16 kHz)	1000	1.87	CC-BY-NC-SA-4.0	MohamedRashad/SADA22
mixat_cs-clean	Emirati + English	podcast	original audio (16 kHz)	1000	3.68	CC-BY-NC-SA-4.0	sqrk/mixat-tri
masc_noisy-clean	mixed	youtube (noisy)	original audio (16 kHz)	1000	0.96	CC-BY-4.0	MohamedRashad/MASC-Arabic
svq_ar-clean	Egyptian, Gulf, Levantine, Maghrebi accents	spoken queries on phones	original audio (16 kHz)	1000	1.24	CC-BY-4.0	google/svq
svq_ar_noisy-clean	Egyptian, Gulf, Levantine, Maghrebi accents	spoken queries on phones in real background speech, media and traffic noise	original audio (16 kHz)	1000	1.29	CC-BY-4.0	google/svq
shams_lev-clean	Levantine (Lebanon, Syria, Palestine, Jordan)	conversations and interviews (LangMedia)	original audio (16 kHz)	907	1.57	CC-BY-SA-4.0	shams-nlp/shams
hassaniya-clean	Hassaniya (Mauritania)	youtube	original audio (16 kHz)	408	1.04	CC-BY-NC-4.0	MauriAI-lab/hassaniya-speech-test-v1
wasil-clean	MSA, Egyptian, Sudanese, Algerian, Syrian	spoken prompts to a chat assistant	original audio (16 kHz)	1000	2.19	CC-BY-NC-SA-4.0	QCRI/WASIL
perle_cs-clean	Egyptian and Saudi Arabic with English	code-switched work and social speech	original audio (16 kHz)	600	1.26	MIT	Perle-ai/ASR_Code_Switch
massive_ar-clean	MSA and Gulf prompts, speakers from across the Arab world	voice-assistant commands	original audio (16 kHz)	1000	1.07	CC-BY-NC-SA-4.0	FBK-MT/Speech-MASSIVE-test
sudanese_sdn-clean	Sudanese	TV programmes	original audio (16 kHz)	1000	1.20	CC-BY-4.0	arbml/sudanese_dialect_speech
fleurs_ar-noisy	MSA	read	+ background noise 0–15 dB SNR (16 kHz)	300	0.91	CC-BY-4.0	google/fleurs
cv18_ar-noisy	MSA	read (crowd-sourced)	+ background noise 0–15 dB SNR (16 kHz)	300	0.36	CC0-1.0	MohamedRashad/common-voice-18-arabic
sada-noisy	Gulf (Saudi)	tv shows	+ background noise 0–15 dB SNR (16 kHz)	300	0.62	CC-BY-NC-SA-4.0	MohamedRashad/SADA22
fleurs_ar-telephony	MSA	read	8 kHz phone channel: band-pass, codec, packet loss	300	0.91	CC-BY-4.0	google/fleurs
cv18_ar-telephony	MSA	read (crowd-sourced)	8 kHz phone channel: band-pass, codec, packet loss	300	0.36	CC0-1.0	MohamedRashad/common-voice-18-arabic
sada-telephony	Gulf (Saudi)	tv shows	8 kHz phone channel: band-pass, codec, packet loss	300	0.62	CC-BY-NC-SA-4.0	MohamedRashad/SADA22
fleurs_ar-callcenter	MSA	read	noise + reverb at the caller, then the phone channel (8 kHz)	300	0.91	CC-BY-4.0	google/fleurs
cv18_ar-callcenter	MSA	read (crowd-sourced)	noise + reverb at the caller, then the phone channel (8 kHz)	300	0.36	CC0-1.0	MohamedRashad/common-voice-18-arabic
sada-callcenter	Gulf (Saudi)	tv shows	noise + reverb at the caller, then the phone channel (8 kHz)	300	0.62	CC-BY-NC-SA-4.0	MohamedRashad/SADA22
fleurs_ar-telephony_v2	MSA	read	phone channel v2: device × network × storage profiles, MENA codec mix, tandem to A-law	300	0.91	CC-BY-4.0	google/fleurs
cv18_ar-telephony_v2	MSA	read (crowd-sourced)	phone channel v2: device × network × storage profiles, MENA codec mix, tandem to A-law	300	0.36	CC0-1.0	MohamedRashad/common-voice-18-arabic
sada-telephony_v2	Gulf (Saudi)	tv shows	phone channel v2: device × network × storage profiles, MENA codec mix, tandem to A-law	300	0.62	CC-BY-NC-SA-4.0	MohamedRashad/SADA22
fleurs_ar-callcenter_v2	MSA	read	caller-side room and environment noise, then the phone channel v2	300	0.91	CC-BY-4.0	google/fleurs
cv18_ar-callcenter_v2	MSA	read (crowd-sourced)	caller-side room and environment noise, then the phone channel v2	300	0.36	CC0-1.0	MohamedRashad/common-voice-18-arabic
sada-callcenter_v2	Gulf (Saudi)	tv shows	caller-side room and environment noise, then the phone channel v2	300	0.62	CC-BY-NC-SA-4.0	MohamedRashad/SADA22
fleurs_ar-sweep	MSA	read	paired SNR grid: 3 noises × 5 SNRs per clip (16 kHz)	2250	7.03	CC-BY-4.0	google/fleurs
masc_clean-sweep	mixed	youtube	paired SNR grid: 3 noises × 5 SNRs per clip (16 kHz)	2250	2.61	CC-BY-4.0	MohamedRashad/MASC-Arabic
fleurs_ar-padded	MSA	read	clips padded with silence, comfort noise, tones or office noise (16 kHz)	100	0.53	CC-BY-4.0	google/fleurs
cv18_ar-padded	MSA	read (crowd-sourced)	clips padded with silence, comfort noise, tones or office noise (16 kHz)	100	0.37	CC0-1.0	MohamedRashad/common-voice-18-arabic
masc_clean-padded	mixed	youtube	clips padded with silence, comfort noise, tones or office noise (16 kHz)	100	0.40	CC-BY-4.0	MohamedRashad/MASC-Arabic
sada-padded	Gulf (Saudi)	tv shows	clips padded with silence, comfort noise, tones or office noise (16 kHz)	100	0.43	CC-BY-NC-SA-4.0	MohamedRashad/SADA22

Licences. Every config keeps the licence of its source dataset. NC = non-commercial only.

Sets not re-hosted

The following sets are part of the benchmark, but their licences do not allow redistribution. Rebuild them locally with scripts/build_benchmark.py; the exact ids are listed in manifests/.

Set	Source	Licence
mgb2	MohamedRashad/mgb2-arabic	QCRI MGB-2 agreement
casablanca	UBC-NLP/Casablanca	CC-BY-NC-ND-4.0
mgb3_egy	ArabicSpeech/MGB-3	unspecified
mgb5_mor	ArabicSpeech/MGB-5	unspecified
algerian_cs	oist/algerian-code-switched-asr-eval	controlled-access terms
nadi_robust	UBC-NLP/NADI2026_subtask1.1_Robust_ASR	unspecified
echonova	echonova-ai/arabic-dialect-asr-benchmark	transcripts CC-BY-4.0; audio research and evaluation only
tedxtn_cs	KBayoud/ASR_TEDx_Tunisie	CC-BY-NC-ND-4.0
mtedx_ar	deepdml/mtedx	CC-BY-NC-ND-4.0
gsb_ar	speechcolab/GigaSpeechBench	unspecified (CC-licensed YouTube audio per the paper)