Electronic apparatus for processing user utterance and controlling method thereof

A system comprising:

• a microphone;
• a speaker;
• a display;
• at least one processor operatively connected to the microphone and the speaker;
• and
• a memory operatively connected to the at least one processor,
• wherein the memory is configured to:
• store an automatic speech recognition (ASR) module, a natural language understanding (NLU) module, and a plurality of response models, and
• store instructions that, when executed, cause the at least one processor to:
• enter into a sleep state or standby state,
• while in the sleep state or the standby state, receive a first voice input including a first wake-up keyword via the microphone, the first wake-up keyword being defined by a wake-up keyword recognition model,
• in response to receiving the first voice input including the first wake-up keyword, exit the sleep state or the standby state and extract first text data corresponding to the first wake-up keyword from the first voice input, using the ASR module,
• select at least one of the response models based at least partly on the extracted first text data corresponding to the first wake-up keyword, and
• display, on the display, a result of executing actions of the selected at least one of the plurality of response models based on the first voice input generated.

A method of processing a voice input, the method comprising:

• entering into a sleep state or standby state;
• while in the sleep state or the standby state, receiving a first voice input including a first wake-up keyword, the first wake-up keyword being defined by a wake-up keyword recognition model;
• in response to receiving the first voice input including the first wake-up keyword, exiting the sleep state or the standby state and extracting first text data corresponding to the first wake-up keyword from the first voice input, using an automatic speech recognition (ASR) module;
• selecting at least one of a plurality of response models based at least partly on the extracted first text data corresponding to the first wake-up keyword;
• and
• displaying a result of executing actions of the selected at least one of the plurality of response models based on the first voice input generated.