Automating communications using an intent classifier

Patent metadata

Field	Value
Publication number	US-12014379-B2
Application number	US-201916597483-A
Country	US
Kind code	B2
Filing date	Oct 9, 2019
Priority date	Dec 8, 2017
Publication date	Jun 18, 2024
Grant date	Jun 18, 2024

Abstract

An intent classifier may be used to increase the efficiency of a communications system. A company may provide assistance to a first user using automated processing or by a second user manually responding to the first user. To reduce costs, the company may prefer to use automated processing for assistance where it is available. While a second user is assisting a user, a message from the first user may be processed with an intent classifier to determine that automated support is available to assist the first user, and a suggestion may be presented to the second user to transfer the first user to the automated processing, such as by presenting a button to the second user to transfer the first user to the automated processing. The second user may then transfer the first user to the automated processing and assist other users.

First claim

What is claimed is:

1. A system for using a neural network for automating responses to user requests, the system comprising:
at least one server comprising at least one processor and at least one memory, the at least one server configured to:
obtain a neural network for processing communications to compute an intent score for each intent of a plurality of possible intents, wherein the plurality of possible intents correspond to categories of user requests;
establish a communications session between a first user and a second user by routing communications through a first API server, a connecting server, and a second API server, wherein the first API server is connected to the first user and the second API server is connected to the second user and wherein the second user is responding to a first request of the first user;
transmit one or more communications between the first user and the second user;
obtain word embeddings for words of the one or more communications, wherein a word embedding for a corresponding word comprises a vector that represents the corresponding word in a vector space;
process the word embeddings with the neural network to compute intent scores;
select a first intent from the plurality of possible intents using the intent scores, wherein the first intent relates to the first request of the first user;
identify a plurality of automated processing tasks that are available to assist the first user with the first request;
receive an indication of a selection of a first automated processing task from the plurality of automated processing tasks by the second user;
change routing of communications of the first user to a first automated processing component through the first API server, thereby transferring the first user from receiving assistance from the second user to receiving assistance with the first automated processing component; and
assist the first user with the first automated processing component.

2. The system of claim 1 , wherein the at least one server is further configured to:
process second word embeddings with the neural network to compute second intent scores, wherein the word embeddings correspond to communications between a third user and a fourth user;
select a second intent from the plurality of possible intents using the second intent scores, wherein the second intent relates to a second request of the third user; and
determine, using the second intent, that automated processing is not available to assist the third user with the second request of the third user.

3. The system of claim 1 , wherein the at least one server is further configured to:
establish a first network connection between a first computer of the first user and a first server computer, wherein the first network connection is a persistent connection, and wherein the persistent connection is used for transmitting the one or more communications and assisting the first user with the first automated processing component.

4. The system of claim 1 , wherein receiving assistance with the first automated processing component comprises transmitting an API request to a server of a company providing a service to the first user.

5. The system of claim 1 , wherein the at least one server is further configured to:
update a database entry to indicate that the first user is receiving assistance from the second user; and
update the database entry to indicate that the first user is receiving assistance using the first automated processing component after receiving the indication of the selection by the second user.

6. The system of claim 1 , wherein the first user is assisted with the first automated processing component without informing the first user of the transfer from receiving assistance from the second user to receiving assistance with the first automated processing component.

7. The system of claim 1 , wherein the at least one server is operated by a third-party company that provides services to a company and the first user is a customer of the company.

8. A computer-implemented method for using a neural network for automating responses to user requests, the method comprising:
obtaining a neural network for processing communications to compute an intent score for each intent of a plurality of possible intents, wherein the plurality of possible intents correspond to categories of user requests;
establishing a communications session between a first user and a second user by routing communications through a first API server, a connecting server, and a second API server, wherein the first API server is connected to the first user and the second API server is connected to the second user and wherein the second user is responding to a first request of the first user;
transmitting one or more communications between the first user and the second user;
obtaining word embeddings for words of the one or more communications, wherein a word embedding for a corresponding word comprises a vector that represents the corresponding word in a vector space;
processing the word embeddings with the neural network to compute intent scores;
selecting a first intent from the plurality of possible intents using the intent scores, wherein the first intent corresponds a plurality of automated processing tasks;
receiving an indication of a selection of a first automated processing task from the plurality of automated processing tasks by the second user; and
changing routing of communications of the first user to a first automated processing component through the first API server, thereby transferring the first user from receiving assistance from the second user to receiving assistance with the first automated processing component.

9. The computer-implemented method of claim 8 , wherein the indication of selection comprises selection using a button in a user interface.

10. The computer-implemented method of claim 8 , wherein the first request of the first user corresponds to updating an address of the first user, processing a payment, changing a PIN of the first user, or changing services received by the first user.

11. The computer-implemented method of claim 8 , comprising:
establishing a first network connection between a first computer of the first user and a first server computer, wherein the first network connection is a persistent connection, and wherein the persistent connection is used for transmitting the one or more communications and assisting the first user with the first automated processing component.

12. The computer-implemented method of claim 11 , wherein the persistent connection comprises a websocket or long polling.

13. The computer-implemented method of claim 8 , comprising:
establishing a first network connection between a first server computer and a first computer of the first user; and
establishing a second network connection between a second server computer and a second computer of the second user;
wherein transmitting the one or more communications comprises:
transmitting the one or more communications via a connecting server, wherein the connecting server allows other servers to publish to channels and to subscribe to the channels;
the first server computer subscribing to a first channel of the connecting server; and
the second server computer subscribing to the first channel of the connecting server.

14. The computer-implemented method of claim 8 , wherein the neural network comprises an intent classifier.

15. The computer-implemented method of claim 8 , comprising:
selecting a second intent from the plurality of possible intents using the intent scores, wherein the second intent c

Assignees

• Asapp Inc

Inventors

• Shaev Jason
• Sehrawat Vicky
• Knaster Rachel
• Wang shang wei
• Sapoznik Gustavo

Classifications

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06F40/284

  Lexical analysis, e.g. tokenisation or collocates · CPC title

• G06F40/216

  using statistical methods · CPC title

• G06F40/56

  Natural language generation · CPC title