A very large speech recorded dataset in Romanian to help the companies to built up automatic speech recognition and text to speech deep neural acoustic models.
The speech dataset is the largest available high quality (recording conditions, -80 dB) voice recorded database containing 61 hours of speech from a large number of users. These, make the dataset a very valuable asset to build automatic speech recognition and text to speech deep neural netowrks acoustic models. The dataset has been practically used in commercial applications for text to speech synthesis and as a prototype for automatic speech recognition in Romanian. Building a chat bot, a virtual assistant or any other human computer interaction system in spoken language in Romanian, obviously would require such dataset. The dataset can not be distributed or sold, as it is subject of Copyright agreements, but it can be successfully used to create acoustic models on the local servers and the acoustic models may be integrated in other business applications.
Automatic transcription of spoken records, text to speech synthesis, chat bots, virtual assistants.
Transcription into text of spoken medical records with an accuracy of 85% at the word level, Multispeaker Text to speech synthesis .