Nü Echo has already collected and transcribed a significant French Canadian telephony acoustic data corpus.

Here is a quick overview of the corpus and the type of licensing that is offered.

Data volume:

• more than 1000 speakers
• adults
• both sexes

Usages:

• Accoustic model training and testing
• Recognition and rejection threshold tuning, etc.

Environment:

• quiet

Type of data:

• Canadian postal codes
• Canadian cities
• Canadian addresses
• Numbers
• Amount of money
• Phone numbers
• Spelling
• Dates
• Times
• Booleans
• Conpany names
• People names

Format:

• Sound files: muLaw, 8 bits
• Transcriptions and annotations: XML files, others possible.
• Phonetic dictionnary: several phone sets supported including SAMPA, UPS, others possible.

Type of license:

• Right to use, royalty free, world-wide, no right to sublicense