FIELD: electronic equipment.
SUBSTANCE: invention relates to means of determining intent associated with uttered utterance. Method includes determining at least one speech element, wherein each speech element has text data representing a word or a pause, and has a corresponding digital audio signal segment. For each speech element the method includes creation of the appropriate text characteristic, creation of the corresponding vector of acoustic characteristics and creation of the corresponding expanded vector of characteristics. Method also includes use of a neural network (NN) configured to determine the intent of the utterance pronounced by the user by inputting the extended characteristic vectors into the NN. NN has been trained to assess the probability that the intention relates to the given type.
EFFECT: taking into account the acoustic characteristics of the utterance uttered by the user to determine the intention associated with the given utterance.
20 cl, 9 dwg
Authors
Dates
2020-01-15—Published
2018-05-23—Filed