What is PENTAtrainer2?

PENTATrainer2 (pen·ta·train·ner·two) is a semi-automatic software packages written as Praat scripts integrated with Java programs. It is developed by Dr Yi Xu (University College London) and Dr Santitham Prom-on (King Mongkut’s University of Technology Thonburi). PENTAtrainer2 facilitates the investigation of underlying representations of communicative functions in any language. Its core concept is based on the Parallel Encoding and Target Approximation (PENTA) framework (Xu, 2005), the quantitative Target Approximation (qTA) model (Prom-on et al., 2009), and the simulated annealing optimization (Kirkpatrick et al., 1983). In contrast to its predecessor, PENTAtrainer1, PENTAtrainer2 globally optimise for the functional parameters of all utterances in the speech corpus simultaneously. Provided with the sound files and their annotations, the program automatically learns the optimal parameters of all possible functional combinations that users have annotated. After the optimization, the learned functional parameters can be used to synthesize F0 contours according to any of the given communicative functions.

graphical_abstract

Why PENTAtrainer2?

  • High synthetic accuracy of prosody
  • Many-to-one mapping from variable surface F0 to invariant underlying targets
  • Effectively handling of both contextual and non-contextual variability
  • Combination of deterministic synthesis and data-driven parameter learning
  • Large-scale and full-detailed prosody synthesis as tool for theory testing

How to cite

Xu, Y. and Prom-on, S. (2014) “Toward invariant functional representations of variable surface fundamental frequency contours: Synthesizing speech melody via model-based stochastic learning”, Speech Communication, 57:181-208.