Granted Patent B2

US 12,591,603 · App. 18/078,740 · Granted Mar 31, 2026

Inventors: Soham Ray (Ithaca, NY); Volkan Cirik (London, GB); Kilian Quirin Weinberger (Ithaca, NY); Satchuthananthavale Rasiah Kuhan Branavan (London, GB); Paloma Sodhi (Ithaca, NY)

Assignee: ASAPP, INC.

G06F16/3329G06F40/30G06F40/40

Abstract

Claims (74)

1 . A computer-implemented method for extracting values from text, comprising:

presenting a user interface to a first user, wherein the user interface comprises a key-value portion and a conversation portion;

receiving a first communication of a conversation from a second user;

processing the first communication to select a first natural language intent;

presenting the first communication in the conversation portion;

selecting a plurality of target keys using the first natural language intent, wherein the plurality of target keys correspond to key values to be extracted from the conversation, and wherein the plurality of target keys includes:

a first target key corresponding to a first key name, a first prompt, and a first value extractor, and

a second target key corresponding to a second key name, a second prompt, and a second value extractor;

presenting the first key name and the second key name in the key-value portion;

initiating an automated process for obtaining, from the second user, a first value for the first target key using the first value extractor and a second value for the second target key using the second value extractor; and

presenting on the user interface an interface element configured to enter, by the first user, an additional target key to be extracted from the conversation.

2 . The computer-implemented method of claim 1 , wherein the automated process comprises:

presenting the first prompt to the second user,

receiving a second communication from the second user,

processing the second communication with the first value extractor to determine the first value,

presenting the first value in the key-value portion,

presenting the second prompt to the second user,

receiving a third communication from the second user, and

processing the third communication with the second value extractor.

3 . The computer-implemented method of claim 1 , wherein the first natural language intent is selected from a set of available natural language intents and each natural language intent may be associated with one or more target keys.

4 . The computer-implemented method of claim 1 , wherein the first value extractor uses a first extraction technique, the second value extractor uses a second extraction technique, and the first extraction technique is different from the second extraction technique.

5 . The computer-implemented method of claim 2 , wherein the automated process includes:

determining that the second value extractor was not able to determine the second value; and

notifying the first user.

6 . The computer-implemented method of claim 2 , comprising restarting the automated process after an action by the first user.

7 . The computer-implemented method of claim 2 , comprising performing an operation on behalf of the second user using the first value and the second value.

8 . The computer-implemented method of claim 2 , wherein the plurality of target keys includes a third target key corresponding to a third key name, and

wherein the automated process includes:

determining that a fourth communication from the second user includes a third value for the third target key, and

presenting the third key name and the third value in the key-value portion.

9 . The computer-implemented method of claim 1 , wherein the automated process is initiated automatically after selecting the plurality of target keys.

10 . A system, comprising at least one server computer comprising at least one processor and at least one memory, the at least one server computer configured to:

receive a first communication of a conversation from a second user to a first user;

process the first communication to select a first natural language intent;

select a plurality of target keys using the first natural language intent, wherein the plurality of target keys correspond to key values to be extracted from the conversation, and wherein the plurality of target keys includes:

a first target key corresponding to a first key name, a first prompt, and a first value extractor, and

a second target key corresponding to a second key name, a second prompt, and a second value extractor; and

present to the first user an interface element for entering an additional target key to be extracted from the conversation.

11 . The system of claim 10 , wherein the at least one server computer is configured to:

provide the first prompt for presentation to the second user;

receive a second communication of the second user;

process the second communication with the first value extractor to determine a first value;

provide the second prompt for presentation to the second user;

receive a third communication of the second user; and

process the third communication with the second value extractor.

12 . The system of claim 10 , wherein the first value extractor uses a first extraction technique, the second value extractor uses a second extraction technique, and the first extraction technique is different from the second extraction technique.

13 . The system of claim 10 , wherein the first value extractor uses a regular expression to determine a first value.

14 . The system of claim 10 , wherein the first value extractor uses a dialogue state tracking to determine a first value.

15 . The system of claim 10 , wherein the first value extractor uses a question-answering model to determine a first value.

16 . The system of claim 11 , wherein the at least one server computer is configured to:

present a first user interface to the first user;

present a second user interface to the second user; and

present the first value in the first user interface.

17 . The system of claim 10 , wherein the first communication is received via an API call from a computer of a company to a computer of a third-party company.

18 . The system of claim 10 , wherein processing the first communication to select a first natural language intent comprises processing the first communication with a neural network classifier.

19 . The system of claim 10 , wherein:

the plurality of target keys includes a third target key corresponding to a third key name; and

retrieving a third value from a datastore.

20 . One or more non-transitory, computer-readable media comprising computer-executable instructions that, when executed, cause at least one processor to perform actions comprising:

receiving a first communication of a conversation from a second user to a first user;

processing the first communication to select a first natural language intent;

selecting a plurality of target keys using the first natural language intent, wherein the plurality of target keys correspond to key values to be extracted from the conversation, and wherein the plurality of target keys includes:

a first target key corresponding to a first key name, a first prompt, and a first value extractor, and

a second target key corresponding to a second key name, a second prompt, and a second value extractor;

providing the first prompt for presentation to the second user;

receiving a second communication of the second user;

processing the second communication with the first value extractor to determine a first value;

providing the second prompt for presentation to the second user;

receiving a third communication of the second user;

processing the third communication with the second value extractor; and

presenting to the first user an interface element for entering an additional target key to be extracted from the conversation.

21 . The one or more non-transitory, computer-readable media of claim 20 , wherein the second user is receiving customer support from a company.

22 . The one or more non-transitory, computer-readable media of claim 20 , wherein the first value extractor uses a first extraction technique, the second value extractor uses a second extraction technique, and the first extraction technique is different from the second extraction technique.

23 . The one or more non-transitory, computer-readable media of claim 20 , wherein the first communication is a text communication.

Assignments (1)

Continuity (1)

References Cited (19)