Mpaligner is the many-to-many string alignment tool based on the generative model which is modified to find a minimum mapping between two strings, such as notation and pronunciation. Mpaligner has some functions. The partial annotation function enables you to give correct alignment to part data by manpower (to provide Semi-supervised training). The detection function of special data detects data that is difficult to do alignment (for example [AAA, tri'plei]). The data which alignment is done is employed as training data. For example, when two strings which alignment is done are notation and pronunciation, it is employed as training data to construct a model for grapheme-to-phoneme conversion (g2p conversion). The license of mpaligner is GNU GPL. If you hope to learn a model with aligned data produced by mpaligner to estimate pronunciation and to estimate a pronunciation with the learned model, please use slearp ( http://sourceforge.jp/projects/slearp/ ) which implements the learning methods for the model and a predict function to estimate a pronunciation. Developer implementing mpaligner is below. NAIST(Nara Institute of Science and Technology)[[BR]] Graduate School of Information Science[[BR]] Augmented Human Communication Laboratory[[BR]] The Doctoral Program[[BR]] Keigo Kubo
This page is a skeleton web page which was automatically set up by OSDN system. The purpose of this page is to encourage the full use of the project web.
You can use the project web area to freely build Web sites to provide project information. For more details on the functions and rules of the project Web and how to use it, please read the documents regarding web site hosting.