Determination of conversation threads in a message channel based on conversational flow and semantic similarity of messages

Patent metadata

Field	Value
Publication number	US-11689488-B2
Application number	US-202117313995-A
Country	US
Kind code	B2
Filing date	May 6, 2021
Priority date	Aug 26, 2019
Publication date	Jun 27, 2023
Grant date	Jun 27, 2023

Abstract

A deep learning module classifies messages received from a plurality of entities into one or more conversation threads. In response to receiving a subsequent message, the deep learning module determines which of the one or more conversation threads and a new conversation thread is contextually a best fit for the subsequent message. The subsequent message is added to the determined conversation thread.

First claim

What is claimed is:

1. A method, comprising:
classifying, by a deep learning module implemented in a computational device, messages received from a plurality of entities into one or more conversation threads;
determining a flow fit measure for a subsequent message for potential addition to each of the one or more conversation threads;
determining a semantic match measure for the subsequent message for the potential addition to each of the one or more conversation threads; and
in response to the determining of the flow fit measure and the semantic match measure, classifying the subsequent message into a conversation thread that has a best flow fit and semantic match combination measure in comparison to other conversation threads of the one or more conversation threads, wherein the flow fit measure represents a fluency of conversation in the conversation thread if the subsequent message is added to the conversation thread.

2. The method of claim 1 , wherein the semantic match measure represents a semantic closeness of conversation in the conversation thread if the subsequent message is added to the conversation thread.

3. The method of claim 1 , wherein elements associated with a message includes a message identifier, an identification of a user who sent the message, and a timestamp of the message.

4. The method of claim 1 , wherein the deep learning module has already classified a plurality of sequentially arriving messages into a plurality of conversation threads at a time the subsequent message arrives.

5. The method of claim 1 , wherein the deep learning module determines whether to start a new conversational thread with the subsequent message, and wherein to start the new conversational thread the subsequent message is added to an empty conversation thread.

6. The method of claim 1 , wherein a flow fit measure is a first conditional probability-based expression, and wherein a semantic match measure is a second conditional probability-based expression.

7. The method of claim 1 , wherein potential conversation threads for adding a new message are pruned by using a greedy search mechanism in which low likelihood conversation threads for adding the new message are discarded based on computed measures that include measures for semantic closeness, measures for fluency of conversation in the conversation thread, and elapsed times of messages.

8. A system, comprising:
a memory; and
a processor coupled to the memory, wherein the processor performs operations, the operations comprising:
classifying, by a deep learning module implemented in a computational device, messages received from a plurality of entities into one or more conversation threads;
determining a flow fit measure for a subsequent message for potential addition to each of the one or more conversation threads;
determining a semantic match measure for the subsequent message for the potential addition to each of the one or more conversation threads; and
in response to the determining of the flow fit measure and the semantic match measure, classifying the subsequent message into a conversation thread that has a best flow fit and semantic match combination measure in comparison to other conversation threads of the one or more conversation threads, wherein the flow fit measure represents a fluency of conversation in the conversation thread if the subsequent message is added to the conversation thread.

9. The system of claim 8 , wherein the semantic match measure represents a semantic closeness of conversation in the conversation thread if the subsequent message is added to the conversation thread.

10. The system of claim 8 , wherein elements associated with a message includes a message identifier, an identification of a user who sent the message, and a timestamp of the message.

11. The system of claim 8 , wherein the deep learning module has already classified a plurality of sequentially arriving messages into a plurality of conversation threads at a time the subsequent message arrives.

12. A computer program product, the computer program product comprising a computer readable storage medium having computer readable program code embodied therewith, the computer readable program code configured to perform operations on a processor, the operations comprising:
classifying, by a deep learning module implemented in a computational device, messages received from a plurality of entities into one or more conversation threads;
determining a flow fit measure for a subsequent message for potential addition to each of the one or more conversation threads;
determining a semantic match measure for the subsequent message for the potential addition to each of the one or more conversation threads; and
in response to the determining of the flow fit measure and the semantic match measure, classifying the subsequent message into a conversation thread that has a best flow fit and semantic match combination measure in comparison to other conversation threads of the one or more conversation threads, wherein the flow fit measure represents a fluency of conversation in the conversation thread if the subsequent message is added to the conversation thread.

13. The computer program product of claim 12 , wherein the semantic match measure represents a semantic closeness of conversation in the conversation thread if the subsequent message is added to the conversation thread.

14. The computer program product of claim 12 , wherein elements associated with a message includes a message identifier, an identification of a user who sent the message, and a timestamp of the message.

Assignees

• IBM

Inventors

• Tan Ming
• Wang Haoyu
• Wang Dakuo
• Gan Chuang

Classifications

• H04L51/216 Primary

  Handling conversation history, e.g. grouping of messages in sessions or threads · CPC title

• G06F40/30

  Semantic analysis · CPC title

• G06F40/35 Primary

  Discourse or dialogue representation · CPC title

• H04L51/42

  Mailbox-related aspects, e.g. synchronisation of mailboxes · CPC title