Database systems with automated structural metadata assignment

Patent metadata

Field	Value
Publication number	US-12248754-B2
Application number	US-202217933385-A
Country	US
Kind code	B2
Filing date	Sep 19, 2022
Priority date	Sep 20, 2021
Publication date	Mar 11, 2025
Grant date	Mar 11, 2025

Abstract

Database systems and methods are provided for assigning structural metadata to records and creating automations using the structural metadata. One method of assigning structural metadata to a record associated with a conversation involves obtaining a plurality of utterances associated with the conversation, identifying, from among the plurality of utterances, a representative utterance for semantic content of the conversation, assigning the conversation to a group of semantically similar conversations based on the representative utterance, and automatically updating the record associated with the conversation at a database system to include metadata identifying the group of semantically similar conversations.

First claim

What is claimed is:

1. A method of assigning structural metadata to a record associated with a conversation, the method comprising:
obtaining a plurality of utterances associated with the conversation, the plurality of utterances including at least a first set of one or more utterances corresponding to a first actor and a second set of one or more utterances corresponding to a second actor;
identifying, from among the plurality of utterances, a representative utterance for semantic content of the conversation from within the first set of one or more utterances of the conversation, wherein the representative utterance comprises a verb and noun indicative of an intent of the first actor that initiated the conversation;
automatically updating a first field of the record associated with the conversation at a database system to include text of the representative utterance from within the conversation;
assigning the conversation to a group of semantically similar conversations from among a plurality of groups of semantically similar conversations based on a relationship between the representative utterance and a reference representative utterance providing a semantic representation of the group of semantically similar conversations; and
automatically updating a second field of the record associated with the conversation at the database system to include metadata identifying the group of semantically similar conversations.

2. The method of claim 1 , wherein:
obtaining the plurality of utterances comprises obtaining the plurality of utterances from a transcript of the conversation; and
the record of the conversation maintains an association between the transcript of the conversation, the first field comprising the text of the representative utterance from within the conversation and the second field comprising the metadata identifying the group of semantically similar conversations.

3. The method of claim 1 , further comprising generating a numerical representation of the representative utterance, wherein assigning the conversation to the group of semantically similar conversations comprises assigning the conversation to the group of semantically similar conversations based on the numerical representation.

4. The method of claim 3 , wherein generating the numerical representation comprises converting content of the representative utterance into a numerical vector representation by inputting the content of the representative utterance into an encoder model.

5. The method of claim 4 , wherein assigning the conversation to the group of semantically similar conversations based on the numerical representation comprises clustering the representative utterance into a cluster group of semantically similar conversations based on a relationship between the numerical vector representation of the representative utterance and one or more numerical vector representations of respective representative utterances associated with respective conversations of the cluster group of semantically similar conversations.

6. The method of claim 5 , wherein the metadata comprises an identifier associated with the cluster group of semantically similar conversations.

7. The method of claim 1 , wherein assigning the conversation to the group of semantically similar conversations based on the representative utterance comprises:
assigning the conversation to a cluster group of semantically similar conversations based on a relationship between the representative utterance and representative utterances associated with respective conversations of the cluster group of semantically similar conversations; assigning the cluster group of semantically similar conversations to a semantic group of a plurality of semantic groups; and
automatically updating a third field of the record to include metadata comprising an identifier associated with the semantic group of the plurality of semantic groups.

8. The method of claim 7 , wherein each semantic group of the plurality of semantic groups is distinct relative to other semantic groups of the plurality of semantic groups.

9. The method of claim 7 , wherein each semantic group of the plurality of semantic groups encompasses a plurality of cluster groups of semantically similar conversations.

10. The method of claim 9 , wherein each cluster group of the plurality of cluster groups of semantically similar conversations is distinct relative to other cluster groups of the plurality of cluster groups.

11. The method of claim 7 , further comprising generating a numerical vector representation of the representative utterance, wherein assigning the conversation to the cluster group of semantically similar conversations comprises assigning the conversation to the cluster group of semantically similar conversations based on a relationship between the numerical vector representation of the representative utterance and one or more numerical vector representations of respective representative utterances associated with respective conversations of the cluster group of semantically similar conversations.

12. The method of claim 11 , wherein assigning the cluster group of semantically similar conversations to the semantic group comprises:
identifying the reference representative utterance for the cluster group of semantically similar conversations;
generating a numerical representation of the reference representative utterance for the cluster group; and
assigning the cluster group to the semantic group based on a relationship between the numerical representation of the reference representative utterance for the cluster group and a second numerical representation of the semantic group.

13. The method of claim 12 , wherein the reference representative utterance for the cluster group comprises a center representative utterance identified from among a plurality of representative utterances associated with respective conversations of the cluster group of semantically similar conversations.

14. The method of claim 12 , wherein the reference representative utterance for the cluster group comprises an autogenerated name associated with the cluster group of semantically similar conversations.

15. The method of claim 1 , wherein:
identifying the representative utterance comprises:
selecting a subset of utterances associated with the first actor, the subset comprising a predetermined number of utterances associated with the first actor; and
sequentially determining, from the subset, an earliest utterance indicative of the intent of the conversation by the first actor; and

the representative utterance comprises the earliest utterance indicative of the intent of the conversation by the first actor.

16. The method of claim 1 , further comprising identifying the reference representative utterance providing the semantic representation of the group of semantically similar conversations as a respective representative utterance for semantic content of a respective conversation of the group of semantically similar conversations corresponding to a numerical vector that represents a center, a median or a mean of the group of semantically similar conversations, wherein the reference representative utterance comprises a respective verb and noun indicative of a respective intent of a third actor that initiated the respective conversation.

17. At least one non-transitory machine-readable storage medium that provides instructions that, when executed by at least one processor, are configurable to cause the at least one processor to:
obtain a plurality of utterances associated with a conversation, the plurality of utt

Assignees

• Salesforce Inc

Inventors

• Mao Yixin
• Alexander Zachary
• Xie Tian
• Liu Wenhao

Classifications

• G10L15/22

  Procedures used during a speech recognition process, e.g. man-machine dialogue · CPC title

• G10L15/20

  Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech (G10L21/02 takes precedence) · CPC title

• G06F16/345

  Summarisation for human users · CPC title

• G06F16/3329

  Natural language query formulation · CPC title

• G10L15/1815

  Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning · CPC title