Contact:
nemlar@hum.ku.dk

BLARK Definition and BLARK Content
- for written language

The degree to which the modules are needed is marked by plus signs:
+++ means essential
++ means very important
+ means important

Speech modules and corresponding spoken language resources, marked with importance:
(Follow the blue link to see the BLARK Content.)

Resources	BNSC	Desktop/Microphone & High quality microphone data or phone data	Telephony	Audio data with prosodic markers and other emotional features	Annotated written corpus	Unannotated written corpus	Vowelised corpus	Non-vowelised corpus	Phonetic lexicon general vocabular	Onomastica (proper names)	Visual data (faces, lips, etc.)
Speech modules
Acoustic models	+++	+++	+++	+++
Language models					++	+++	++	+++
Pronunciation lexicon					+		++		+++	+++
Lexicon adaptation					+	+	++	+	+++	+++
Phoneme alignment	++	++	++	++	++		+		+++	+++
Prosody recognition				++	++		++		++	++
Speech Units Selection		+	+	+++	++				+	+
Prosody prediction				++	++		++		++	++
Segmentation Speech / Silence	++	++	++	++
Sentence boundary detection	++	++	++	++					+	+
Dialect / language identification	++	++	++	+					+	+
Word boundary identification	+	+	+	+					+	+
Speech /Non-speech (music) detection	++	+	+	++
Speaker recognition / identification	+	+	+	+
Emotion identification	+	+	+	+					+	+
Speaker adaptation	++	++	++	+
Lips movement reading											+++

Speech applications and corresponding speech modules, marked with importance:

Applications	Dictation	Telephony speech applications	Embedded speech recognition	Transcription of broadcast News	Transcription of conversational speech	Speaker recognition	Dialect / language identification	Emotion identification	Speaker adaptation	Lips movement reading	Topic detection, segmentation, topic boundaries	Speaker 2 speaker mapping	Emotion / Prosody output	Text to Speech (incl. formatted data e.g. databases)	Customization to different voices	Generation Lips Movement
Speech modules
Acoustic models	+++	+++	+++	+++	+++	++	+++	+++	+++	+++	+++	++	+++	+++	+++	+++
Language models	+++	++	++	+++	+++		++						++	+++
Pronunciation lexicon	+++	+++	+++	+++	+++							++		+++
Lexicon adaptation	+	+	+	+	+							++		+++
Phoneme alignment	+	+	+	+	+	+	++					++		+++
Prosody recognition	+	+	+	+	+	+	+	+++	+			++
Speech Units Selection													+++	+++
Prosody prediction													+++	+++
Segmentation Speech / Silence	++	+	++	++	++	+	++	++	+	+	+		+
Sentence boundary detection	+	+	+	+	+	+	+	++	+	+	+		++	+++
Dialect / language identification	+	+	+	+	+	+	+	+	+	+	+			+
Word boundary identification	+	+	+	+	+	+	+	+	+	+	+		++
Speech /Non-speech (music) detection	+	+	+	+	+	+	+	++	+	+	+
Speaker recognition / identification
Emotion identification	+	+	+	+	+	+	+		+	+	+	++	++
Speaker adaptation	++	+	++	+	++	+	+	+	+	+	+	++		+
Lips movement reading										+++

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