New-word pronunciation learning using a pronunciation graph

Method and computer - readable medium conversion voice to describe the character the user pronunciation of text and character, and voice identifying on dictionary. The start, comprising at least two multiple possibility voice descriptions. Represents the character the voice signal user pronunciation to form a voice to describe with the decoding. With at least one voice descriptions of the character's text. A based on language and text-based voice rendering multiple possibility sequences, based on connected to the user pronunciation's homologue, and a to application and digital of the score of the single pattern. And and voice composed of high score and magnetic the voice identifying dictionary.