Method and system for conversation detection in email systems
View Patent ↗A method and system are provided for conversation detection in email systems. Multiple email messages are provided and grouped as relating to a conversation. The grouping is carried out by applying a similarity function based on a similarity of the email messages' attributes, the similarity function including a similarity between the email messages' participants and at least one of a similarity between the email messages' subjects or a similarity between the email messages' contents. The similarity function may also include the similarity between the email messages' dates. The similarity function may also include weightings for the contributions of the email messages' attributes. A graphical user interface is provided in an email client which includes means for viewing email messages by conversation.
1. A method for conversation detection in email systems, comprising:
providing a computer system, wherein the system comprises distinct program modules embodied on a computer-readable medium, and wherein the distinct program modules comprise a conversation detector;
executing the conversation detector to group email messages into first groups, each of the first groups having a common core subject to thereby define a conversation by applying a similarity function a first time based on a similarity of the email messages' attributes, the similarity function including:
a similarity between the email messages' participants;
at least one of a similarity between the email messages' subjects and a similarity between the email messages' contents;
applying the similarity function a second time to group the email messages of the first groups into respective second groups; and
defining the second groups as sub-conversations.
2. The method as claimed in claim 1 , wherein the similarity function also includes a similarity between the email messages' dates.
3. The method as claimed in claim 2 , including defining a maximum date difference between email messages, above which the date similarity is zeroed.
4. The method as claimed in claim 1 , wherein the similarity function includes weightings for the contributions of the email messages' attributes.
5. The method as claimed in claim 4 , wherein the weightings are optimized using machine learning.
6. The method as claimed in claim 1 , further including:
applying the similarity function a third time to group the sub-conversations function into new groups; and
defining the new groups as new conversations.
7. The method as claimed in claim 6 , including sorting the sub-conversations by starting date prior to applying the similarity function a third time.
8. The method as claimed in claim 1 , including sorting the email messages by date prior to applying the similarity function a second time.
9. The method as claimed in claim 1 , including comparing a new message to conversations that previous messages have been grouped into, detecting an optimal conversation for a new message to be added to.
10. The method as claimed in claim 1 , wherein the similarity between the email messages' participants includes distinguishing between active and passive participants by applying an activity weighting to a participant in an email message.
11. A system for conversation detection in email systems, comprising:
a processor; and
a memory accessible to the processor storing programs and data objects therein, the objects including a mailbox containing a plurality of email messages each having defined attributes, wherein execution of the programs cause the processor to group the email messages to define a conversation by applying a similarity function a first time based on a similarity of the email messages' attributes, the similarity function including: a similarity between the email messages' participants; and at least one of a similarity between the email messages' subjects and a similarity between the email messages' contents, wherein the processor is operative for:
grouping the email messages together in subject groups having a same core subject;
grouping the email messages within the subject groups by applying the similarity function a second time, and defining first resultant groups as sub-conversations; and
grouping the sub-conversations across all the subject groups by applying the similarity function a third time, and defining second resultant groups as new conversations.
12. The system as claimed in claim 11 , wherein the similarity function also includes a similarity between the email messages' dates.
13. The system as claimed in claim 11 , means wherein the processor is operative to set weightings for contributions of the email messages' attributes in the similarity function.
14. The system as claimed in claim 11 , wherein the processor is operative for comparing a new message to conversations that previous messages have been grouped into, and for detecting an optimal conversation for addition of the new message to thereto.
15. An email client system for conversation detection, comprising:
a processor; and
a memory accessible to the processor storing programs and data objects therein, the objects including a mailbox containing a plurality of email messages each having defined attributes
a display; and
wherein the programs comprise a conversation detector and a graphical user interface for displaying the email messages, and the processor is operative to execute the conversation detector to detect a conversation among the email messages
by applying a similarity function a first time based on a similarity of the email messages' attributes, the similarity function including: a similarity between the email messages' participants; and at least one of a similarity between the email messages' subjects and a similarity between the email messages' contents and to present an indication of the conversation on the display using the graphical user interface, wherein the processor is operative for:
grouping the email messages together in subject groups having a same core subject;
grouping the email messages within the subject groups by applying the similarity function a second time, and defining first resultant groups as sub-conversations; and
grouping the sub-conversations across all the subject groups by applying the similarity function a third time, and defining second resultant groups as new conversations.
16. The email client system as claimed in claim 15 , wherein the processor is operative for adding the email messages to the conversation and removing the email messages a from the conversation.
17. The email client system as claimed in claim 15 , means wherein the processor is operative to select one of a plurality of conversation choices for a new email message to be added, the conversation choices being best similarity matches to the new email message from conversations comprising previous email messages.
18. A computer program product stored on a computer readable storage medium for conversation detection in email systems, comprising computer readable program code for instructing a computer to perform the steps of:
grouping email messages into a conversation by applying a similarity function a first time based on a similarity of the email messages' attributes, the similarity function including: a similarity between the email messages' participants; and at least one of a similarity between the email messages' subjects and a similarity between the email messages' contents;
grouping the email messages together in subject groups having a same core subject;
grouping the email messages within the subject groups by applying the similarity function a second time, and defining first resultant groups as sub-conversations; and
grouping the sub-conversations across all the subject groups by applying the similarity function a third time, and defining second resultant groups as new conversations.