Folders, labels and alerts may some day be supplanted by algorithms that keep users' inboxes organized. A University College Dublin professor is at work on just such a programmatic technique to recognize unstructured processes in e-mail messages and use that to generate specific workflows.
Working with IBM Research, Nicholas Kushmerick has applied algorithms in a system known as automated e-mail activity management. This system sorts e-mails automatically, without user input.
Once messages are sorted, additional algorithms are used to organize e-mails based on their related activities, place the messages in the proper order and then present them as a single workflow. "In one end of the pipeline goes your inbox. Out the other end of the pipeline comes a structure that is automatically discovered and senses a pattern," Kushmerick said.
For the last year, Kushmerick has been a visiting scientist at IBM's Dublin Software Laboratory and has worked with Tessa Lau of IBM's T.J. Watson Research Center. Now Kushmerick is working with IBM product and research groups focused on the next generation of e-mail. Patent applications were filed in several countries for the algorithm technology at the heart of scheme to automatically sort e-mail according to use.
In January, Kushmerick and Lau published their research, which applied the algorithms to the e-mail trail accompanying online purchases. However, Kushmerick sees the application being more useful for large enterprises that must organize e-mail exchanges related to planning meetings, reimbursing employee expenses or confirming travel plans.
"You presumably are scheduling lots of different meetings with lots of different people simultaneously," Kushmerick said. "There's a sequence of messages that is easy to lose. Within the enterprise, there could be dozens or even hundreds of these workflows."
The algorithms were successful, meaning they ranked an e-mail in the right spot in the workflow 91 percent of the time, according to the authors' research.
Kushmerick suggested an appropriate user interface could help in workflow supervision, because users may not want the system to sort through personal or confidential e-mails.
"The real hard work in turning this idea into a product is not so much building more software but building the business case," he said. "What will the end user get out of this? Is it a 'gee whiz' program or will it save time and money?"
Kushmerick hopes the e-mail management technology can be ready for commercial use in two to three years.