I work with speech recognition for non-native speakers of English, especially
those in India. One of the major challenges with this user group is their
variable and non-native accent, which means that I should manually update the
pronunciation dictionary based on some common pronunciations. However, does
anyone know of a source code that could automate this process, even offline,
i.e. not real-time, but a source code that picks up audio files from a
directory, and a mapping of the words that they represent, and generate the
updated pronunciation dictionary?
Hi,
I work with speech recognition for non-native speakers of English, especially
those in India. One of the major challenges with this user group is their
variable and non-native accent, which means that I should manually update the
pronunciation dictionary based on some common pronunciations. However, does
anyone know of a source code that could automate this process, even offline,
i.e. not real-time, but a source code that picks up audio files from a
directory, and a mapping of the words that they represent, and generate the
updated pronunciation dictionary?
There are some papers that have already talked about this approach, such as:
http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=04430113, but I'm wondering if there
already open-source source codes that people have written for this task?
Thanks very much,