METHOD, DEVICE, AND ARTICLE OF MANUFACTURE FOR THE GENERATION BASED ON A NEURONAL NETWORK OF POST-LEXIC PRONOUNCIATIONS FROM LEXICAL PRONUNCIATIONS

摘要

A method (2000), device (2200), and article of manufacture (2300) provide, in response to lexical pronunciation information, cost-effective generation of post-lexical pronunciation information. A method is presented for ensuring, in response to lexical pronunciation, the cost-effective generation of post-lexical pronunciation, including the following steps: determining lexical phonemes, lexical characteristics, and information on boundaries for a portion predetermined text; and using a preformed neural network which has been preformed using lexical phonemes, post-lexical phonemes, lexical characteristics, and boundary information, to generate a neural network hypothesis for post-lexical pronunciation of the predetermined portion of text.