Method and system for processing unclear intent query in conversation system

1. A query processing method comprising:

inputting a query into an input of a user device, the query being based on an utterance of a user and a context associated with the user;

transmitting the query via a network to at least one processor from the user device;

analyzing the query for an intent of the user by the at least one processor based on a natural language understanding (NLU) for the query;

predicting the intent of the user by the at least one processor by using a deep learning based probabilistic model having the query and the context as an input when the intent of the user is undetermined from the analyzing of the query;

providing a response from the at least one processor to the user device based on the predicted intent of the user and the context associated with the user; and

executing a function by the at least one processor when the user confirms that the predicted intent associated with the response is a correct intent of the user.

2. The method of claim 1 , wherein the context includes information about at least a portion of a dialog history associated with the user and a status for each of a plurality of items of a device of the user receiving the utterance of the user.

3. The method of claim 1 , wherein the predicting of the intent of the user comprises generating an entire representation for an intent prediction by concatenating at least two of a Dialog Flow Representation including a transition sequence of a dialog state acquired based on a dialog history included in the context, a Device Status Representation including a status vector about statuses of a device further included in the context, and a Sentence Representation including a tokenized input sentence for the query, and predicting, as the intent of the user, an intent that is determined based on a statistical probability corresponding to the entire representation through the deep learning based probabilistic model.

4. The method of claim 3 , wherein the Dialog Flow Representation includes a state transition sequence of the dialog history acquired through a neural net based classification model that is trained to determine a state of an input dialog.

5. The method of claim 3 , wherein the Device Status Representation includes a status vector of the device acquired using a neural net based classification model that is trained to output one of statuses preset for a specific item based on status information of the specific item.

6. The method of claim 3 , wherein the Sentence Representation includes a tokenized input sentence for the query acquired using a neural net based classification model that is trained to tokenize an input sentence and to output a sequence of tokenized texts.

7. The method of claim 1 , wherein the providing of the response comprises:

specifying a function based on the predicted intent, and generating and forwarding a confirmation response for receiving a confirmation from the user on execution of the specified function or a guide response for providing a guide about a method of using the specified function.

8. The method of claim 1 , wherein the providing of the response comprises generating the response by further using a partial analysis result acquired based on the NLU for the query.

9. The method of claim 8 , wherein the partial analysis result includes information about a target of the predicted intent.

10. The method of claim 1 , wherein the analyzing of the intent of the user comprises analyzing the intent of the user by further using a dialog state management based on the context.

11. A non-transitory computer-readable record medium storing computer instructions that, when executed by a processor, cause the processor to perform the query processing method of claim 1 .

12. A computer apparatus comprising:

at least one processor configured to execute computer-readable instructions,

wherein the at least one processor is configured to:

receive a query from a user device, the query being based on an utterance of a user and a context associated with the user,

analyze the query for an intent of the user based on a natural language understanding (NLU) for the query,

predict the intent of the user using a deep learning based probabilistic model having the query and the context as an input when the intent of the user is undetermined from the analyzing of the query,

provide a response to the user device based on the predicted intent of the user and the context associated with the user, and

execute a function when the user confirms that the predicted intent associated with the response is a correct intent of the user.

13. The computer apparatus of claim 12 , wherein the context includes information about at least a portion of a dialog history associated with the user and a status for each of a plurality of items of a device of the user receiving the utterance of the user.

14. The computer apparatus of claim 13 , wherein the at least one processor is further configured to

generate an entire representation for an intent prediction by connecting at least two of a Dialog Flow Representation including a transition sequence of a dialog state acquired based on a dialog history included in the context, a Device Status Representation including a status vector about states of a device further included in the context, and a Sentence Representation including a tokenized input sentence for the query, and

predict, as the intent of the user, an intent that is determined based on a statistical probability corresponding to the entire representation through the deep learning based probabilistic model.

15. The computer apparatus of claim 14 , wherein the Dialog Flow Representation includes a state transition sequence of the dialog history acquired through a neural net based classification model that is trained to determine a state of an input dialog.

16. The computer apparatus of claim 14 , wherein the Device Status Representation includes a status vector of the device acquired using a neural net based classification model that is trained to output one of statuses preset for a specific item based on status information of the specific item.

17. The computer apparatus of claim 14 , wherein the Sentence Representation includes a tokenized input sentence for the query acquired using a neural net based classification model that is trained to tokenize an input sentence and to output a sequence of tokenized texts.

18. The method of claim 1 , wherein providing the response includes providing a confirmation response requesting confirmation by the user that the predicted intent is the correct intent.

19. The computer apparatus of claim 12 , wherein the response includes providing a confirmation response requesting confirmation by the user that the predicted intent is the correct intent.