Electronic device and method for providing conversational service

Patent metadata

Field	Value
Publication number	US-11710481-B2
Application number	US-202016797339-A
Country	US
Kind code	B2
Filing date	Feb 21, 2020
Priority date	Aug 26, 2019
Publication date	Jul 25, 2023
Grant date	Jul 25, 2023

Abstract

A method, performed by an electronic device, of providing a conversational service includes: receiving an utterance input; identifying a temporal expression representing a time in a text obtained from the utterance input; determining a time point related to the utterance input based on the temporal expression; selecting a database corresponding to the determined time point from among a plurality of databases storing information about a conversation history of a user using the conversational service; interpreting the text based on information about the conversation history of the user, the conversation history information being acquired from the selected database; generating a response message to the utterance input based on a result of the interpreting; and outputting the generated response message.

First claim

What is claimed is:

1. A method, performed by an electronic device, of providing a conversational service to a user, the method comprising:
receiving an utterance input;
identifying a temporal expression representing a time in a text obtained from the utterance input by applying a pre-trained temporal expression extraction model to the obtained text;
determining a time point related to the utterance input based on the temporal expression, wherein the determined time point related to the utterance input corresponds to a time before the utterance input is received;
selecting a database corresponding to the determined time point from among a plurality of databases storing information about a conversation history of the user using the conversational service;
interpreting the text based on information about the conversation history of a user, the conversation history information being acquired from the selected database;
generating a response message to the utterance input based on a result of the interpreting; and
outputting the generated response message,
wherein the plurality of databases comprise a first database storing information about the conversation history of the user accumulated before a preset time point and a second database storing information about the conversation history of the user accumulated after the preset time point, and
wherein the selecting of the database corresponding to the determined time point from among the plurality of databases comprises:
selecting the first database from among the plurality of databases based on the determined time point related to the utterance input being before the preset time point; and
selecting the second database from among the plurality of databases based on the determined time point related to the utterance input being after the preset time point, and
wherein the determining of the time point related to the utterance input comprises:
predicting probability values including probabilities that the temporal expression will represent each of a plurality of time points; and
determining a time point, corresponding to a highest probability value from among the predicted probability values, as the time point related to the utterance input.

2. The method of claim 1 , wherein the identifying of the temporal expression comprises:
obtaining the text by performing speech recognition on the utterance input; and
determining, as the temporal expression, an entity representing at least one of a time point included in the text, a duration included in the text, or a period included in the text.

3. The method of claim 2 , wherein the determining of the entity comprises:
performing embedding for mapping the text to a plurality of vectors;
assigning a beginning-inside-outside (BIO) tag to at least one morpheme representing at least one of the time point, the duration, or the period included in the text by applying a bidirectional long short-term memory (LSTM) model to the plurality of vectors; and
identifying the entity in the text based on the BIO tag.

4. The method of claim 1 , further comprising:
based on ending of a conversation service during which the utterance input is received, transmitting to the first database information about a user's conversation history accumulated in the second database while the conversation service was provided.

5. The method of claim 1 , wherein the first database is stored in an external server, and the second database is stored in the electronic device, and
wherein the preset time point includes one of a time point based on at least some of the information about the conversation history of the user, included in the second database, being transmitted to the first database, a time point based on a face image of the user being obtained, and a time point based on the conversational service starting.

6. The method of claim 1 , wherein the interpreting of the text comprises:
determining an entity included in the text that needs to be specified;
acquiring specification information for specifying the determined entity by retrieving the information about the conversation history of the user, acquired from the selected database; and
interpreting the text and the specification information using a natural language understanding (NLU) model.

7. The method of claim 1 , wherein the generating of the response message comprises:
determining a type of the response message by applying a dialog manager (DM) model to the result of the interpreting; and
generating the response message of the determined type using a natural language generation (NLG) model.

8. The method of claim 1 , further comprising:
obtaining a face image of the user;
determining whether a face ID corresponding to the obtained face image is stored by searching the first database included in the plurality of databases; and
initiating the conversational service based on a result of the determining.

9. The method of claim 8 , wherein the initiating of the conversational service comprises:
based on the face ID corresponding to the obtained face image being stored in the first database, updating a stored service usage history mapped to the face ID; and
based on the face ID corresponding to the obtained face image not being stored in the first database, generating a new face ID and a service usage history mapped to the new face ID.

10. The method of claim 8 , further comprising:
transmitting the face ID to another electronic device after the conversational service ends; and
transmitting, in response to a request received from the other electronic device, information about the conversation history of the user stored in the second database included in the plurality of databases, to the other electronic device.

11. An electronic device configured to provide a conversational service to a user, the electronic device comprising:
a memory storing one or more instructions; and
at least one processor configured to execute the one or more instructions to provide the conversational service to the user,
wherein the at least one processor is further configured to execute the one or more instructions to control the electronic device to:
receive an utterance input; identify a temporal expression representing a time in a text obtained from the utterance input by applying a pre-trained temporal expression extraction model to the obtained text; determine a time point related to the utterance input based on the temporal expression, wherein the determined time point related to the utterance input corresponds to a time before the utterance input is received; select a database corresponding to the determined time point from among a plurality of databases storing information about a conversation history of the user using the conversational service; interpret the text based on information about the conversation history of the user, the conversation history information being acquired from the selected database; generate a response message to the utterance input based on a result of the interpreting; and output the generated response message,
wherein the plurality of databases comprise a first database storing information about the conversation history of the user accumulated before a preset time point and a second database storing information about the conversation history of the user accumulated after the preset time point, and
wherein the selecting of the database corresponding to the determined time point from among the plurality of databases comprises:
selecting the first database from among the plurality of databases based on the determined time point related to the utterance input being before the preset time point; and
selecting the second databas

Assignees

• Samsung Electronics Co Ltd

Inventors

• Ham Jina
• Lee Kangwook
• Kim Soofeel
• Park Yewon
• Choi Wonjong

Classifications

• G10L15/22 Primary

  Procedures used during a speech recognition process, e.g. man-machine dialogue · CPC title

• G06F16/3329 Primary

  Natural language query formulation · CPC title

• G10L2015/223

  Execution procedure of a spoken command · CPC title

• G10L2015/025

  Phonemes, fenemes or fenones being the recognition units · CPC title

• H04L51/02

  using automatic reactions or user delegation, e.g. automatic replies or chatbot-generated messages · CPC title