Secure complete phrase utterance recommendation system

Patent metadata

Field	Value
Publication number	US-12288032-B2
Application number	US-202318499077-A
Country	US
Kind code	B2
Filing date	Oct 31, 2023
Priority date	Nov 18, 2019
Publication date	Apr 29, 2025
Grant date	Apr 29, 2025

Abstract

Described herein are systems, apparatus, methods and computer program products for machine learning intent classification. In various embodiments, historical utterances provided by users may be utilized for bot training. Context and personally identifiable information may be removed from the utterances. The utterances may be associated with vectors. The utterances and vectors may be used to determine recommendations.

First claim

The invention claimed is:

1. A database system, comprising:
an utterance database configured to store utterance data associated with one or more utterance phrases;
a phrase vector database configured to store phrase vector data comprising one or more phrase vectors, wherein each phrase vector is associated with a corresponding utterance phrase within the utterance database; and
a processor configured to perform operations comprising:
receiving an utterance dataset comprising a plurality of training phrases;
decoupling each of the training phrases from the other of the training phrases;
creating training data by:
associating each of the training phrases with phrase unique identifiers;
associating phrase vectors with each of the training phrases; and
associating each of the phrase vectors with phrase vector unique identifiers, wherein each phrase vector unique identifier is matched with an associated phrase unique identifier;

training a model with the training data;
receiving entry data;
determining, with the model, one or more phrase vectors matching the entry data;
determining, based on the determined one or more phrase vectors, one or more utterance phrases associated with the one or more phrase vectors; and
communicating the one or more utterance phrases to a user device for display on a graphical user interface of the user device.

2. The database system of claim 1 , wherein the operations further comprise:
determining, based on the model, an entry vector associated with the entry data.

3. The database system of claim 1 , wherein the one or more utterance phrases are determined as recommendations by the model.

4. The database system of claim 1 , wherein the determining, with the model, the one or more phrase vectors matching the entry data comprises matching one or more phrase vectors within the phrase vector database to the entry data.

5. The database system of claim 1 , wherein the determining the one or more utterance phrases comprises requesting the utterances phrases associated with the determined one or more phrase vectors from the utterance database.

6. The database system of claim 5 , wherein the determining the one or more utterance phrases comprises providing authentication data.

7. The database system of claim 1 , wherein the operations further comprise:
decoupling each of the training phrases from the other of the training phrases; and
removing personally identifiable information from each of the training phrases.

8. The database system of claim 7 , wherein the removing the personally identifiable information comprises:
analyzing the utterance dataset to identify words present less than a threshold number of times within the utterance dataset; and
removing the identified words.

9. The database system of claim 7 , wherein the removing the personally identifiable information comprises:
analyzing the utterance dataset to identify words present in conversations between a number of parties less than a threshold number; and
removing the identified words.

10. The database system of claim 7 , wherein the utterance database receives the utterance dataset.

11. A method comprising:
receiving an utterance dataset comprising a plurality of training phrases;
decoupling each of the training phrases from the other of the training phrases;
creating training data by:
associating each of the training phrases with phrase unique identifiers;
associating phrase vectors with each of the training phrases; and
associating each of the phrase vectors with phrase vector unique identifiers, wherein each phrase vector unique identifier is matched with an associated phrase unique identifier;

training a model with the training data;
receiving entry data;
determining, with the model, one or more phrase vectors matching the entry data, wherein the one or more phrase vectors are each associated with a corresponding utterance phrase stored within an utterance database, and wherein the utterance database is configured to store utterance data associated with the one or more utterance phrases;
determining, based on the determined one or more phrase vectors, one or more utterance phrases associated with the one or more phrase vectors; and
communicating the one or more utterance phrases to a user device for display on a graphical user interface of the user device.

12. The method of claim 11 , further comprising:
determining, based on the model, an entry vector associated with the entry data.

13. The method of claim 11 , wherein the one or more utterance phrases are determined as recommendations by the model.

14. The method of claim 11 , wherein the determining with the model, the one or more phrase vectors matching the entry data comprises matching one or more phrase vectors within a phrase vector database to the entry data.

15. The method of claim 11 , wherein the determining the one or more utterance phrases comprises requesting the utterances phrases associated with the determined one or more phrase vectors from the utterance database.

16. The method of claim 15 , wherein the determining the one or more utterance phrases comprises providing authentication data.

17. The method of claim 11 , further comprising:
decoupling each of the training phrases from the other of the training phrases; and
removing personally identifiable information from each of the training phrases.

18. The method of claim 17 , wherein the removing the personally identifiable information comprises:
analyzing the utterance dataset to identify words present less than a threshold number of times within the utterance dataset; and
removing the identified words.

19. The method of claim 17 , wherein the removing the personally identifiable information comprises:
analyzing the utterance dataset to identify words present in conversations between a number of parties less than a threshold number; and
removing the identified words.

20. The method of claim 17 , wherein the utterance database receives the utterance dataset.

Assignees

• Salesforce Inc

Inventors

• Kale Anuprit
• Peng Weiping
• Cheng Na
• Lindstrom Rick
• Alexander Zachary

Classifications

• G06F16/3344

  using natural language analysis · CPC title

• G06F16/3329

  Natural language query formulation · CPC title

• G06F16/31

  Indexing; Data structures therefor; Storage structures · CPC title

• G06F40/30 Primary

  Semantic analysis · CPC title

• G06F16/3347

  using vector based model · CPC title