Bermuda, a data-driven tool for phonetic transcription of words

The article presents the Bermuda component of the NLPUF text-to-speech toolbox. Bermuda performs phonetic transcription for out-of-vocabulary words using a Maximum Entropy classifier and a custom designed algorithm named DLOPS. It offers direct transcription by using either one of the two available algorithms, or it can chain either algorithm to a second layer Maximum Entropy classifier designed to correct the first-layer transcription errors. Bermuda can be used outside of the NLPUF package by itself or to improve performance of other modular text-to-speech packages. The training steps are presented, the process of transcription is exemplified and an initial evaluation is performed. The article closes with usage examples of Bermuda.