Work to improve Watson Text to Speech (TTS) service. Collaborate with teammates to create new test fixtures and enhance Voice Tuner to expedite development.
•Established Brazilian Portuguese tokenizer (BPT), analyzing existing tokenizer codebase and implementing BPT in Watson TTS.
•Improved tokenizer accuracy by creating new test fixtures using Python Behave Framework to capture wrong phoneme output during regression testing.
•Streamlined voice tuning process by using C++ REST API, enabling Voice Tuner application to communicate with a server using JSON to synthesize speech from input text, request information regarding the sample recordings of produced speech, and edit voice properties to tune voices based on customer requirements.
•Fixed memory corruption in Japanese speech synthesis engine using gdb and valgrind.