Artificial intelligence apparatus and method for recognizing speech in consideration of utterance style

1. An artificial intelligence apparatus for recognizing speech in consideration of an utterance style, comprising:

a microphone;

a memory configured to store a plurality of speech recognition models corresponding to a plurality of utterance styles, respectively; and

a processor configured to:

obtain, via the microphone, speech data including speech of a user,

extract an utterance feature vector from the obtained speech data,

determine an utterance style corresponding to the speech based on the extracted utterance feature vector and an utterance style determination model,

determine whether the determined utterance style corresponding to the speech is a new utterance style,

based on a determination that the determined utterance style corresponding to the speech is not the new utterance style, generate a first speech recognition result by using a speech recognition model corresponding to the determined utterance style, the speech recognition model having been selected from among the plurality of speech recognition models stored in the memory,

based on a determination that the determined utterance style corresponding to the speech is the new utterance style, generate a new speech recognition model and generate a second speech recognition result by using the generated new speech recognition model,

wherein the processor is configured to:

map the extracted utterance feature vector to an utterance feature space,

determine a cluster closest to the mapped utterance feature vector among clusters corresponding to a plurality of previously learned utterance styles,

determine the utterance style corresponding to the extracted utterance feature vector as an utterance style corresponding to the closest cluster based on a distance from the mapped utterance feature vector to the closest cluster being less than a predetermined reference value, the utterance style corresponding to the closest cluster being stored in the memory, and

determine the utterance style corresponding to the utterance feature vector as a new utterance style based on the distance being equal to or greater than the predetermined reference value, wherein the utterance style corresponding to the closest cluster is not one of the plurality of speech recognition models stored in the memory.

2. The artificial intelligence apparatus of claim 1 ,

wherein the speech recognition model is learned using training data including speech data corresponding to an utterance style corresponding to the speech recognition model.

3. The artificial intelligence apparatus of claim 1 ,

wherein the first or second speech recognition result includes intent information corresponding to the speech.

4. The artificial intelligence apparatus of claim 1 ,

wherein the utterance style determination model includes an artificial neural network and is learned using a machine learning algorithm or a deep learning algorithm.

5. The artificial intelligence apparatus of claim 1 ,

wherein the processor is configured to:

generate training data corresponding to the extracted utterance feature vector if the determined utterance style is a new utterance style,

learn the new speech recognition model corresponding to the new utterance style using the generated training data, and

generate the second speech recognition result corresponding to the speech using the learned new speech recognition model.

6. The artificial intelligence apparatus of claim 5 ,

wherein the processor is configured to:

generate speech data corresponding to the extracted utterance feature vector from a predetermined text set using a text-to-speech (TTS) engine, and

generate training data including the predetermined text set and the generated speech data.

7. The artificial intelligence apparatus of claim 1 ,

wherein the utterance feature vector includes at least one of a gender of a speaker, a speech speed, a pronunciation, a pronunciation stress, a pause interval, a pitch, a tone, an intonation, a rhyme or an emotion.

8. A method of recognizing speech in consideration of an utterance style, the method comprising:

obtaining, via a microphone, speech data including speech of a user;

extracting an utterance feature vector from the obtained speech data;

determining an utterance style corresponding to the speech based on the extracted utterance feature vector and an utterance style determination model;

determining whether the determined utterance style corresponding to the speech is a new utterance style;

based on a determination that the determined utterance style corresponding to the speech is not the new utterance style, generating a first speech recognition result by using a speech recognition model corresponding to the determined utterance style, the speech recognition model having been selected from among a plurality of speech recognition models stored in a memory;

based on a determination that the determined utterance style corresponding to the speech is the new utterance style, generating a new speech recognition model and generating a second speech recognition result by using the generated new speech recognition model,

wherein the memory is configured to store the plurality of speech recognition models corresponding to a plurality of utterance styles, respectively, and

wherein determining whether the determined utterance style corresponding to the speech is the new utterance style comprises:

mapping the extracted utterance feature vector to an utterance feature space,

determining a cluster closest to the mapped utterance feature vector among clusters corresponding to a plurality of previously learned utterance styles,

determining the utterance style corresponding to the extracted utterance feature vector as an utterance style corresponding to the closest cluster based on a distance from the mapped utterance feature vector to the closest cluster being less than a predetermined reference value, the utterance style corresponding to the closest cluster being stored in the memory, and

determining the utterance style corresponding to the utterance feature vector as a new utterance style based on the distance being equal to or greater than the predetermined reference value, wherein the utterance style corresponding to the closest cluster is not one of the plurality of speech recognition models stored in the memory.

9. A non-transitory computer-readable medium having recorded thereon a program for performing a method of recognizing speech in consideration of an utterance style, the method comprising:

obtaining, via a microphone, speech data including speech of a user;

extracting an utterance feature vector from the obtained speech data;

determining an utterance style corresponding to the speech based on the extracted utterance feature vector and an utterance style determination model;

determining whether the determined utterance style corresponding to the speech is a new utterance style;

based on a determination that the determined utterance style corresponding to the speech is not the new utterance style, generating a first speech recognition result by using a speech recognition model corresponding to the determined utterance style, the speech recognition model having been selected from among a plurality of speech recognition models stored in a memory;

based on a determination that the determined utterance style corresponding to the speech is the new utterance style, generating a new speech recognition model and generating a second speech recognition result by using the generated new speech recognition model,

wherein the memory is configured to store the plurality of speech recognition models corresponding to a plurality of utterance styles, respectively, and

wherein determining whether the determined utterance style corresponding to the speech is the new utterance style comprises:

mapping the extracted utterance feature vector to an utterance feature space,

determining a cluster closest to the mapped utterance feature vector among clusters corresponding to a plurality of previously learned utterance styles,

determining the utterance style corresponding to the extracted utterance feature vector as an utterance style corresponding to the closest cluster based on a distance from the mapped utterance feature vector to the closest cluster being less than a predetermined reference value, the utterance style corresponding to the closest cluster being stored in the memory, and

determining the utterance style corresponding to the utterance feature vector as a new utterance style based on the distance being equal to or greater than the predetermined reference value, wherein the utterance style corresponding to the closest cluster is not one of the plurality of speech recognition models stored in the memory.