Information processing apparatus, keyword detecting apparatus, and information processing method

An information processing method, performed by an information processing apparatus, the method comprising:

• acquiring first training data, including at least one combination of a voice feature quantity and a correct phoneme label of the voice feature quantity;
• training, using the first training data, an acoustic model in a manner to output the correct phoneme label in response to input of the voice feature quantity, the input being performed by the acoustic model, the acoustic model outputting the correct phoneme label;
• extracting, from the first training data, second training data including voice feature quantities of data pieces and correcting phoneme labels of the voice feature quantities, the data pieces including a preset keyword and at least one of a sub-word included in the preset keyword, a syllable included in the preset keyword, or a phoneme included in the preset keyword;
• and
• adapting the trained acoustic model using the second training data to a keyword model used for detection of the preset keyword, the keyword model being generated from the trained acoustic model, the keyword model adapted for the preset keyword being retrained based on the trained acoustic model, wherein
• the extracting extracts second data pieces as the second training data up to a predetermined number of data pieces, in descending order according to a proportion in number of a letter of the preset keyword, a letter of the sub-word, the syllable, or the phoneme to a data piece.

A non-transitory computer readable medium including computer executable instructions, wherein the instructions, when executed by a processor, cause the processor to perform a method comprising:

• acquiring first training data including at least one combination of a voice feature quantity and a correct phoneme label of the voice feature quantity;
• training, using the first training data, an acoustic model in a manner to output the correct phoneme label in response to input of the voice feature quantity, the input being performed by the acoustic model, the acoustic model outputting the correct phoneme label;
• extracting, from the first training data, second training data including voice feature quantities of data pieces and correcting phoneme labels of the voice feature quantities, the data pieces including a preset keyword and at least one of a sub-word included in the preset keyword, a syllable included in the preset keyword, or a phoneme included in the preset keyword;
• and
• adapting the trained acoustic model using the second training data to a keyword model used for detection of the preset keyword, the keyword model being generated from the trained acoustic model, the keyword model adapted for the preset keyword being retrained based on the trained acoustic model, wherein
• the extracting extracts second data pieces as the second training data up to a predetermined number of data pieces, in descending order according to a proportion in number of a letter of the preset keyword, a letter of the sub-word, the syllable, or the phoneme to a data piece.