Granted Patent B1

US 12,244,556 · App. 18/112,442 · Granted Mar 4, 2025

Inventors: Ying Jiang (Fremont, CA); Apoorv Sharma (San Mateo, CA); Brian Matthew Holligan (San Francisco, CA)

Assignee: Doma Technology LLC

H04L51/48G06F40/205G06F40/30G06N20/00G06V30/19173H04L51/063H04L51/08

Abstract

Claims (86)

1. A method comprising:

receiving a message;

parsing the message to identify individual sentences within the message;

encoding each sentence into a vector that represents a semantic meaning of the sentence;

applying one or more classification models to the message including: providing each vector to a machine learning model, wherein the machine learning model is trained to generate a prediction of one or more classifications for each sentence of the message, the classification for each sentence of the message indicating a prediction of whether the sentence is actionable or not, wherein a classification of actionable indicates that the sentence is associated with a process of a real estate transaction and includes content for advancing a workflow of the real estate transaction;

generating a predicted classification for the message as a whole as being actionable or unactionable based on the classifications of individual sentences of the message; and

determining a message routing based on the predicted classification.

2. The method of claim 1 , wherein generating a predicted classification for the message as a whole comprises:

determining whether at least one sentence has a first predicted classification; and

responsive to determining that at least one sentence has the first predicted classification, assigning the message with the first classification.

3. The method of claim 1 , wherein applying one or more classification models comprises:

applying a rule-based classifier to the message; and

overriding a classification of the machine learning model in response to a conflicting classification by the rule-based classifier.

4. The method of claim 3 , wherein applying the rule-based classifier to the message comprises:

determining whether the message includes an attachment of a specified file type; and

responsive to determining that the message includes the attachment of the specified file type, assigning a first classification to the message.

5. The method of claim 3 , wherein applying the rule-based classifier to the message comprises:

applying a domain knowledge base to the message content; and

in response to determining that the message includes one or more terms from the knowledge base, assigning a first classification to the message.

6. The method of claim 1 , further comprising:

prior to parsing the message, performing one or more preprocessing operations to remove extraneous content from the message.

7. The method of claim 6 , wherein the one or more preprocessing operations comprise removing text associated with other messages in a message chain.

8. The method of claim 1 , wherein determining the message routing comprises:

for a first predicted classification of the message, routing the message to one or more human operators or system processes to further process the message; and

for a second predicted classification of the message, routing the message to a default storage location.

9. A system comprising:

one or more computers and one or more storage devices on which are stored instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:

receiving a message;

parsing the message to identify individual sentences within the message;

encoding each sentence into a vector that represents a semantic meaning of the sentence;

applying one or more classification models to the message including: providing each vector to a machine learning model, wherein the machine learning model is trained to generate a prediction of one or more classifications for each sentence of the message, the classification for each sentence of the message indicating a prediction of whether the sentence is actionable or not, wherein a classification of actionable indicates that the sentence is associated with a process of a real estate transaction and includes content for advancing a workflow of the real estate transaction;

generating a predicted classification for the message as a whole as being actionable or unactionable based on the classifications of individual sentences of the message; and

determining a message routing based on the predicted classification.

10. The system of claim 9 , wherein generating a predicted classification for the message as a whole comprises:

determining whether at least one sentence has a first predicted classification; and

responsive to determining that at least one sentence has the first predicted classification, assigning the message with the first classification.

11. The system of claim 9 , wherein applying one or more classification models comprises:

applying a rule-based classifier to the message; and

overriding a classification of the machine learning model in response to a conflicting classification by the rule-based classifier.

12. The system of claim 11 , wherein applying the rule-based classifier to the message comprises:

determining whether the message includes an attachment of a specified file type; and

responsive to determining that the message includes the attachment of the specified file type, assigning a first classification to the message.

13. The system of claim 11 , wherein applying the rule-based classifier to the message comprises:

applying a domain knowledge base to the message content; and

in response to determining that the message includes one or more terms from the knowledge base, assigning a first classification to the message.

14. The system of claim 9 , wherein the instructions are further operable to cause the one or more computers to perform operation comprising:

prior to parsing the message, performing one or more preprocessing operations to remove extraneous content from the message.

15. The system of claim 14 , wherein the one or more preprocessing operations comprise removing text associated with other messages in a message chain.

16. The system of claim 9 , wherein determining the message routing comprises:

for a first predicted classification of the message, routing the message to one or more human operators or system processes to further process the message; and

for a second predicted classification of the message, routing the message to a default storage location.

17. One or more computer-readable storage media encoded with instructions that, when executed by one or more computers, cause the one or more computers to perform operations comprising:

receiving a message;

parsing the message to identify individual sentences within the message;

encoding each sentence into a vector that represents a semantic meaning of the sentence;

applying one or more classification models to the message including: providing each vector to a machine learning model, wherein the machine learning model is trained to generate a prediction of one or more classifications for each sentence of the message, the classification for each sentence of the message indicating a prediction of whether the sentence is actionable or not, wherein a classification of actionable indicates that the sentence is associated with a process of a Page real estate transaction and includes content for advancing a workflow of the real estate transaction;

generating a predicted classification for the message as a whole as being actionable or unactionable based on the classifications of individual sentences of the message; and

determining a message routing based on the predicted classification.

18. The computer-readable storage media of claim 17 , wherein generating a predicted classification for the message as a whole comprises:

determining whether at least one sentence has a first predicted classification; and

responsive to determining that at least one sentence has the first predicted classification, assigning the message with the first classification.

19. The computer-readable storage media of claim 17 , wherein applying one or more classification models comprises:

applying a rule-based classifier to the message; and

overriding a classification of the machine learning model in response to a conflicting classification by the rule-based classifier.

20. The computer-readable storage media of claim 19 , wherein applying the rule-based classifier to the message comprises:

determining whether the message includes an attachment of a specified file type; and

responsive to determining that the message includes the attachment of the specified file type, assigning a first classification to the message.

21. The computer-readable storage media of claim 19 , wherein applying the rule-based classifier to the message comprises:

applying a domain knowledge base to the message content; and

in response to determining that the message includes one or more terms from the knowledge base, assigning a first classification to the message.

22. The computer-readable storage media of claim 17 , wherein the instructions further cause the one or more computers to perform operations comprising:

prior to parsing the message, performing one or more preprocessing operations to remove extraneous content from the message.

23. The computer-readable storage media of claim 22 , wherein the one or more preprocessing operations comprise removing text associated with other messages in a message chain.

24. The computer-readable storage media of claim 17 , wherein determining the message routing comprises:

for a first predicted classification of the message, routing the message to one or more human operators or system processes to further process the message; and

for a second predicted classification of the message, routing the message to a default storage location.

25. A method comprising:

receiving a message;

parsing the message to identify individual sentences within the message;

encoding each sentence into a vector that represents a semantic meaning of the sentence;

applying a machine learning model to the message including: providing each vector to the machine learning model, wherein the machine learning model is trained to generate a prediction of one or more classifications for each sentence of the message, the classification for each sentence of the message indicating a prediction of whether the sentence is actionable or not, wherein a classification of actionable indicates that the sentence is associated with a process of a real estate transaction and includes content for advancing a workflow of the real estate transaction;

generating a predicted first classification for the message as a whole as being actionable or unactionable based on the machine learning model classifications of individual sentences of the message;

applying a rule based classifier to the message including applying one or more rules to the message that define a message as actionable or not;

generating a second classification for the message as a whole as being actionable or unactionable based on whether one or more of the rules are satisfied;

determining a message classification based on the first classification for the message and the second classification for the message; and

determining a message routing based on the determined message classification.

Assignments (4)

Continuity (1)

References Cited (45)