Contact:
nemlar@hum.ku.dk

Spoken Resources

Availability:
3 existent but only company-internal, 2 existent and freely usable for PreR&D, 1 existent and freely usable for both PreR&D and R&D.
Cost:
4 > 10,000 €, 3 1,000 - 10,000 €, 2 100 - 1,000 €, 1 < 100 € or free
Adaptability:
3 black box, 2 glass box (you can see but not change it) 1 freely manipulable

R means for research, C means for commercial use.
For availability = 3 (company internal) other features are irrelevant.

)

Name of Resource	Provider	Size	Other information	Availability, cost, manip.
Acustic Data
SpeechDat like database	UOB/ENS		More than 100 speakers French/Arabic, For speech recognition, Lebanese/Syrian/French	1,1,1
Arabic digits	UOB		For speech recognition, Lebanese accent	1,1,1
Speech database in 4 languages	LibanCell	10,000 announcement with 10 words/announcements		3
Labelled database for TTS	Millenium			3
Arabic broadcast news speech corpus (BNSC)	ELRA/LDC	More than 20 hours of transcribed Arabic news in Modern Standard Arabic.	Domain: news	1,2,1
Arabic acoustic corpus mono-speaker	Benabbou, Morocco			3
Arabic Phonetic database	King Abdulaziz City for Science and Technology		Lang: En-Ar	3
Holy Quran multi-speaker	RDI	60 hours		1,4,1
Single male speaker concatenative Arabic TTS database	RDI	1 hour, 1,300 sentences		1,3,1
Single female speaker concatenative Arabic TTS database	RDI	4 hours, 3,000 sentences		1,3,1
Arabic concatenative TTS male recording	Sakhr	MSA 3 hours		3
Arabic ASR recording db	Sakhr	56 hours of MSA and Colloquial Arabic		3
Human Names Language Model	Sakhr	500,000 Names	Egyptian and Saudi human names corpus	3
Arabic Acoustic Model	Sakhr			3
CALLHOME Egyptian Arabic Speech	LDC	120 Egyptian Colloquial Arabic telephone conversations	Calls lasting up to 30 minutes	1,2,1
CALLFRIEND Egyptian Arabic	LDC	60 telephone conversations between native speaker of Egyptian dialect of Arabic	Calls lasted between 5 and 30 minutes. Includes documentation. All calls are domestic.	1,2,1
CALLHOME Egyptian Arabic Speech Supplement	LDC	20 telephone conversations. Transcripts for 120 Egyptian Colloquial Arabic telephone conversations. 273,681,144 bytes (261 Mbytes) or 8 hours of audio data.	20 data files in sphere format, 8 KHz shorten-compressed 2-channel mulaw.	1,1,1
GlobalPhone Arabic	ELRA	About 100 adult native speakers were asked to read 100 sentences.	The GlobalPhone corpus provides transcribed speech data for the development and evaluation of large vocabulary continuous speech recognition systems.	1,3
OrienTel United Arab Emirates MSA	ELRA	500 speakers (254 males, 246 females)	Recorded over the local fixed and mobile telephone network.	1,4
OrienTel Arabic as spoken in Israel	ELRA	750 Arabic speakers (375 males, 375 females)	Recorded over the Israeli fixed and mobile telephone network.	1,4
OrienTel Jordan MCA	ELRA	757 Jordanian speakers (393 males, 364 females)	Recorded over the Jordanian fixed and mobile telephone network.	1,4
OrienTel Jordan MSA	ELRA	556 Jordanian speakers (288 males, 268 females)	Recorded over the Jordanian fixed and mobile telephone network.	1,4
OrienTel Egypt MCA	ELRA	750 Egyptian speakers (398 males, 352 females)	Recorded over the Egyptian fixed and mobile telephone network.	1,4
OrienTel Egypt MSA	ELRA	500 Egyptian speakers (254 males, 246 females)	Recorded over the Egyptian fixed and mobile telephone network.	1,4
OrienTel Morocco MCA	ELRA	772 Moroccan speakers (383 males, 389 females)	Recorded over the Moroccan fixed and mobile telephone network.	1,4
OrienTel Morocco MSA	ELRA	530 Moroccan speakers (264 males, 266 females)	Recorded over the Moroccan fixed and mobile telephone network.	1,4
OrienTel Tunisia MCA	ELRA	792 Tunisian speakers (426 males, 366 females)	Recorded over the Tunisian fixed and mobile telephone network.	1,4
OrienTel Tunisia MSA	ELRA	598 Tunisian speakers (359 males, 239 females)	Recorded over the Tunisian fixed and mobile telephone network.	1,4
OrienTel United Arab Emirates MCA	ELRA	880 speakers (432 males, 448 females) v	Recorded over the local fixed and mobile telephone network.	1,4
Arabic Broadcast news	LDC		Recordings from several Arabic radio channels	$700
The Corpus of Spoken Palestinian Arabic (CoSPA	University of Haifa, Israel	Between 1996 and 1998, 200 hours of recorded speech have been collected.	The aim is to collect data that would cover the whole linguistic area of Palestinian Arabic.	-
KACST Arabic Phonetics Database	KACST, Saudi Arabia	The database contains more than 46,000 files.	The KAPD is a detailed and comprehensive database that shows the articulatory mechanism of Arabic sounds.	KAPD is available on 3 CDs for researchers and students of Speech.
Saudi Accented Arabic Voice Bank	KACST, Saudi Arabia	1033 native speakers	Saudi accented Arabic telephone speech database	Can be licensed to be used in research or to develop products when a contract with KACST is signed.

Name of Resource	Provider	Size	Other information	Availability, cost, manip.
Written Corpus for Speech Technologys
Corpus for di-syllables	Abdelhak Mouradi, Noureddine Chenfour		Domain: text-to-speech	1,2,1
CALLHOME Egyptian Arabic Transcripts	LDC	Contiguous 5 or 10 minute segments taken from 120 unscripted telephone conversations	The transcripts are timestamped by speaker turn for alignment with the speech signal and are provided in standard orthography.	1,2,1

Name of Resource	Provider	Size	Other information	Availability, cost, manip.
Phonetic Lexicon
Special pronunciations dictionary	Sakhr	20,000 entries	Dict. for handling pronunciation anormalities	3
Name master dictionary	Sakhr	100,000 Names		3
LC-STAR Standard Arabic Phonetic lexicon	ELRA	110,271 entries	52,981 common word entries, 50,135 proper names, 7,155 special application words.	1,4

MEDAR is supported by the European Commission's ICT programme and is running from
February 1st 2008 until July 31st 2010