70K words, Non-validated sentence segmentation. Non-validated POS tagging, Manual annotation of syntactic dependencies and dependency labels, Manual annotation of semantic roles, Manual annotation of events based on a shallow domain specific ontology (only for a 31K words subset of GDT)
Collection of telephone databases from mediterranean region, incl. (variants of) Arabic. 500-1000 speakers per database, all orthographically transcribed. Speaker information regarding gender, age and accent. Phonetic lexicons included.
9 speech databases for training and testing multilingual speech recognition applications in the car environment. Contains parallel 4 channel in-car recordings and a GSM channel. Contains interesting phonetically rich material. All orthographically transcribed. Speaker information included for gender, age, accent. Including pronunciation lexicon.