Interaction method and apparatus based on artificial intelligence

Patent metadata

Field	Value
Publication number	US-10664504-B2
Application number	US-201815891258-A
Country	US
Kind code	B2
Filing date	Feb 7, 2018
Priority date	May 22, 2017
Publication date	May 26, 2020
Grant date	May 26, 2020

Abstract

The present disclosure discloses an interaction method and apparatus based on artificial intelligence. A specific embodiment of the method comprises: receiving a current interactive statement entered by a user through a terminal; extracting at least one type of characteristic based on the current interactive statement and a previous interactive statement entered by the user; processing the at least one type of characteristic using a pretrained neural network model to determine whether an intent maintaining relationship exists between the current interactive statement and the previous interactive statement; and if the intent maintaining relationship exists, updating a limitation condition for the previous interactive statement using the current interactive statement, performing information retrieval using the previous interactive statement with the updated limitation condition, and pushing a retrieved retrieval result to the terminal. This embodiment reduces time spent in user entry.

First claim

What is claimed is:

1. An interaction method based on artificial intelligence, the method comprising:
receiving a current interactive statement entered by a user through a terminal;
extracting at least one type of characteristic based on the current interactive statement and a previous interactive statement entered by the user;
processing the at least one type of characteristic using a pretrained neural network model to determine whether an intent maintaining relationship exists between the current interactive statement and the previous interactive statement, the neural network model being configured to characterize whether a corresponding relationship of the intent maintaining relationship exists between a characteristic extracted from the current and previous interactive statements, and the current and previous interactive statements; and
if the intent maintaining relationship exists, updating a limitation condition for the previous interactive statement using the current interactive statement, performing information retrieval using the previous interactive statement with the updated limitation condition, and pushing a retrieved retrieval result to the terminal.

2. The method according to claim 1 , wherein the extracting at least one type of characteristic based on the current interactive statement and a previous interactive statement entered by the user comprises at least one of:
performing a word segmentation operation on the current interactive statement, and extracting a word obtained by the word segmentation as the characteristic;
segmenting at least one independent interactive substatement present in a user search record out of the current interactive statement using a forward maximum matching approach or a backward maximum matching approach, querying from the user search record respectively using each interactive substatement to obtain a webpage title of a search result clicked by the user among search results obtained by searching by the user using the interactive substatement, performing a word segmentation on the webpage title and extracting a word obtained by the word segmentation as the characteristic;
extracting the characteristic according to whether an intent-related word matching an intent of the previous interactive statement is present in the current interactive statement; and
extracting the characteristic according to whether a statement combination composed of a statement fragment in the current interactive statement and the intent-related word in the previous interactive statement is present in a search statement of the user search record.

3. The method according to claim 1 , wherein the processing the at least one type of characteristic using a pretrained neural network model to determine whether an intent maintaining relationship exists between the current interactive statement and the previous interactive statement comprises:
for each type of characteristic among the at least one type of characteristic, converting each characteristic among the type of characteristic into a corresponding characteristic vector, and performing at least one of summation, convolution and maxpooling on each characteristic vector to generate a first fixed-length vector corresponding to the type of characteristic;
connecting the first fixed-length vector corresponding to each type of characteristic to obtain a second fixed-length vector;
inputting the second fixed-length vector into a SoftMax classifier; and
determining whether the intent maintaining relationship exists between the current interactive statement and the previous interactive statement based on a classification result outputted by the classifier.

4. The method according to claim 3 , wherein the extracting at least one type of characteristic based on the current interactive statement and the previous interactive statement entered by the user further comprises:
extracting a word in the previous interactive statement as the characteristic; and
the processing the at least one type of characteristic using a pretrained neural network model to determine whether an intent maintaining relationship exists between the current interactive statement and the previous interactive statement further comprises:
mapping the word in the previous interactive statement to a word vector, and converting the mapped word vector into a third fixed-length vector using a recurrent neural networks model; and
the inputting the connected second fixed-length vector into a SoftMax classifier comprises:
connecting and then inputting the second fixed-length vector and the third fixed-length vector into the SoftMax classifier.

5. The method according to claim 1 , further comprising:
determining whether an intent switching relationship exists between the current interactive statement and the previous interactive statement if the intent maintaining relationship does not exist; and
selecting the retrieval result corresponding to the previous interactive statement as an object specific to the current interactive statement and operating the object based on the current interactive statement if the intent switching relationship exists.

6. The method according to claim 5 , wherein the operating the object based on the current interactive statement comprises:
when the current interactive statement is a query statement, querying an object property of the object based on the current interactive statement, and pushing the queried object property to the terminal.

7. The method according to claim 5 , wherein the operating the object based on the current interactive statement comprises:
when the current interactive statement is a feedback statement, generating feedback information fed back by the user regarding the object based on the current interactive statement.

8. An interaction apparatus based on artificial intelligence, the apparatus comprising:
at least one processor; and
a memory storing instructions, which when executed by the at least one processor, cause the at least one processor to perform operations, the operations comprising:
receiving a current interactive statement entered by a user through a terminal;
extracting at least one type of characteristic based on the current interactive statement and a previous interactive statement entered by the user;
processing the at least one type of characteristic using a pretrained neural network model to determine whether an intent maintaining relationship exists between the current interactive statement and the previous interactive statement, the neural network model being configured to characterize whether a corresponding relationship of the intent maintaining relationship exists between a characteristic extracted from the current and previous interactive statements and the current and previous interactive statements; and
updating, if the intent maintaining relationship exists, a limitation condition for the previous interactive statement using the current interactive statement, performing information retrieval using the previous interactive statement with the updated limitation condition, and pushing a retrieved retrieval result to the terminal.

9. The apparatus according to claim 8 , wherein the extracting at least one type of characteristic based on the current interactive statement and a previous interactive statement entered by the user comprises at least one of:
performing a word segmentation operation on the current interactive statement, and extracting a word obtained by the word segmentation as the characteristic;
segmenting at least one independent interactive substatement present in a user search record out of the current interactive statement using a forward maximum matching approach or a backward maximum matching approach, que

Assignees

• Beijing Baidu Netcom Sci & Tec

Inventors

• Chen Liwei
• Zhou Xiao
• Yu Dianhai
• Zhao Shiqi

Classifications

• G06F16/3325 Primary

  Reformulation based on results of preceding query · CPC title

• H04L51/02

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

• G06N3/02

  Neural networks · CPC title

• G06F16/335 Primary

  Filtering based on additional data, e.g. user or group profiles (filtering in web context G06F16/9535, G06F16/9536) · CPC title

• G06F16/3329

  Natural language query formulation · CPC title