Estimating the time until a reply email will be received using a recipient behavior model

• US 9,146,895 B2
• Filed: 09/26/2012
• Issued: 09/29/2015
• Est. Priority Date: 09/26/2012
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Lynar, Timothy M., Steer, Kent C.
• Primary Examiner(s): Chaki, Kakali
• Assistant Examiner(s): Seck, Ababacar
• Application Number:
  US13/627,075
• Publication Number:
  US 20140089242A1
• Time in Patent Office: 1,098 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: H04L 51/02 using automatic reactions or user delegation, e.g. automatic replies or chatbot-generated messages
• CPC Class Codes: H04L 51/42 Mailbox-related aspects, e.g. synchronisation of mailboxes

Abstract

A method is provided in an email communication system having at least a processor. The method includes generating an estimate of a receipt time of a reply email from a recipient of an initial email using a recipient behavior model that applies machine learning to the initial email and to training data from other emails. The method further includes indicating the estimate to a user using an indication device.

Claims

1. A method in an email communication system having at least a processor, the method comprising:

sharing a set of recipient behavior models among a set of trusted peers;

generating an estimate of a receipt time of a reply email from a recipient of an initial email using more than one recipient behavior model from the set of recipient behavior models that applies machine learning to the initial email and to training data from other emails; and

indicating the estimate to a user using an indication device,

wherein the set of recipient behavior models are configured in a master-slave configuration, and a master recipient behavior model from among the recipient behavior models in the set is used to output the estimate that is indicated to the user.

2. The method of claim 1, wherein the training data comprises at least one of a date sent, a sender, co-recipients, a recipient calendar, sender-receiver relationship information, message attributes, and a number of unread emails in an email inbox of the recipient.

3. The method of claim 1, wherein the machine learning comprises at least one of neural networks and decision tree induction.

4. The method of claim 1, further comprising generating an updated estimate based on at least one of additional training data that is newly acquired and evaluated and the same training data being further evaluated over time.

5. The method of claim 4, wherein the updated estimate is generated responsive to a new recipient being added or an existing recipient being removed from the initial email.

6. The method of claim 4, wherein the updated estimate is generated responsive to a user request for the updated estimate.

7. The method of claim 1, wherein the training data is obtained from previous personal correspondence between a sender of the initial email and the recipient of the initial email.

8. The method of claim 1, wherein the training data is obtained from previous peer correspondence between trusted peers, the trusted peers including the recipient of the initial email.

9. The method of claim 1, wherein the training data represents a subset of data corresponding to the previous peer correspondence, the subset determined to preserve private details in the training data.

10. The method of claim 1, wherein at least one previous actual reply time is used to generate the estimate.

11. The method of claim 1, further comprising generating a confidence measure of the estimate.

12. The method of claim 1, wherein said sharing step comprises receiving at least one of the recipient behavior models in the set from a trusted peer, and wherein the more than one recipient behavior model used to generate the estimate comprises the at least one of the recipient behavior models received from the trusted peer.

13. The method of claim 1, wherein the recipient behavior model models a particular recipient from among a plurality of possible recipients, and the method further comprises selecting the recipient behavior model from among a plurality of recipient behavior models that include at least different two recipient behavior models for the recipient of the initial email.

14. The method of claim 1, wherein the training data comprises a number of unread emails in an email inbox of the recipient.

15. The method claim 1, wherein the estimate is updated responsive to any keystroke while a message body has focus.