Email is a structured form of communication that is usually not isolated but part of much a much larger construct. Although email programs provide tools like folder and follow-up options, most of the world's users choose to allow their inboxes to remain cluttered.
Researchers discovered that by using email's basic structure, all messages can be organized by task. University College and IBM researchers from Dublin, Ireland responsible for the study determined that most emails are not random but are connected to some purpose whether management of travel, scheduling or business dealings. By using researcher's techniques, users can organize their messages and structure them into recognizable, easy to find information.
This technique could lead to development of task scheduling tools that will support the organization of email inboxes while at the same time allow for search capability within the business.
The researchers developed a three-prong prototype to implement this new strategy.
- All similar messages are grouped together by task. For example, an eBay auctioneer would group various types of auctions into separate categories i.e. clothing, shoes, jewelry.
- The task software scans the entire spectrum of activity and then performs regrouping of messages based on that information. Back to the eBay auctioneer example - all eBay messages would be scanned and placed in the proper category together i.e. "thanks for bidding" message would go in the appropriate clothing, shoe or jewelry group identified in #1 above.
- Finally, the software will combine all organization into one model that will demonstrate the process order as it occurs. Because real-world tasks can often loop, a single auction might have multiple "outbid" emails sent out so it can become complicated when trying to order the components.
The classification and clustering of text also included using induction algorithms to carry out figure out word meanings and define sets of information. The algorithms themselves construct flow models and while each of these methods are normally isolated, no previous researchers ever integrated these methodologies together.
Although this is a shallow approach compared to other models normally used for representing data, the systems the research study utilized are so structured that this 3 step process works. For example, an item number classifies each eBay auction so the system could take this information and organize message using that unique identification.
The system's automation is so user friendly that a person could give a set of instructions, tell it to organize and the system can do it…all without asking for any further information.
Next on researcher's agenda is to test in the real world using larger data collections. They are also improving the system's interface to make it easier to correct mistakes and help the system learn how to take data and generalize it. This constant improving of the system is necessary since full automation will never be totally accurate when applied to extremely complex situations. To date, the prototype is working is working at 91 percent accuracy in the classification and organization of messages.
When the computer algorithm is unable to complete a task, researchers are seeking solutions for the computer to be able to ask users for clarification in order to create the structure. Once this is accomplished, users would be able to query the computer regarding other sorts of activities, such as "calculate my online purchases last year?" and "what are my medical expenses this year?"
Researchers are also making progress on using this technique for person to person messages rather than just machine to person.
It will be 2 or 3 years before this technology is available commercially. IBM has filed for tech patents and is presently looking at commercial feasibility.
Eventually, this technology could lead to automation tools such as email scheduling and analysis features. For example, you send a message and request a document from the recipient. This would be added to a pending project line that would be marked satisfied once the recipient emails the document to you.
Ultimately it would mean that the average computer user would be able to completely customize their computer experience through the use of software and tools with fully customizable features.
