Contacts out of date? ‘Semantic’ e-mail could allow you to address with less stress
Imagine that you have just drawn the short straw at work and must now plan a surprise birthday party for your division’s longtime vice president. The invite list, you’re told, should include not only all the other vice presidents and the division’s current employees, but also all the employees who used to work in your division but are now elsewhere in the company. You could do hours of research to compile the e-mail list, or, if your company is running a Stanford-developed technology called SEAMail, you could send that e-mail before your coffee gets cold.
SEAMail, developed by a team led by computer science Associate Professor Michael Genesereth, is a prototype, Web-based e-mail system that allows users to address messages merely by describing the intended recipients (e.g., “vice presidents” or “engineering employees”). For users, the technology obviates the need to discover or remember exact e-mail addresses. In the party scenario, you wouldn’t have to know a single recipient’s name, much less e-mail address, to invite the right people.
Being a professor, Genesereth imagines sending an e-mail to all CS students whose previous courses might make them eligible to take a new class. SEAMail would, in principle, make it easy to address all the students who’ve taken the right prerequisite classes. “Currently, it would be a rather burdensome task for me to find out who those people are, but it is very easy for me to describe who should get the message,” he says.
This summer, the entire computer science department — a community of more than 600 core users and thousands more casually associated ones — will get the chance to try the system out. This will be an important test of SEAMail, which has been used before only within a small community of Genesereth’s colleagues at other institutions, mostly in Europe. That trial, and the technology itself, are described in a paper published in the January edition of the journal IEEE Internet Computing. Genesereth’s co-authors are graduate student Michael Kassoff, senior research scientist Charles Petrie and former student Lee-Ming Zen, now at Microsoft.
Finding meaning among data
The SEA in SEAMail stands for Semantic E-mail Addressing. Semantics, in the context of electronic data, refers to the idea that computers should be able to “understand” data based on the meaning it has for users. SEAMail will, for example, be able to integrate the following facts (and many more) from three administrative databases in the CS department: Michael Genesereth is an associate professor, with the office number Gates 220, and his research concerns computational logic.
Sure, one could build a database to support a query for those facts, but SEAMail doesn’t require that step. The data integration technology under the hood, software called Infomaster, is built to integrate data from multiple sources whether or not they were intended to work together and whether or not they were built for making e-mail addressing easier. Endowing computers with the capacity to interpret and integrate structured data from multiple sources has been a major thrust of Genesereth’s work for decades.
Genesereth is quick to caution that no matter how “smart” the software is, the system will only be as capable as the supply of data allows it to be. For this reason, he says, the technology will likely find its greatest use within organizations such as companies, who have well-maintained directories and other personnel databases.
Ultimately, however, the advent of the “Semantic Web” could extend SEAMail’s reach world wide. The Semantic Web, whose biggest advocates include Web inventor Tim Berners-Lee, specifies a set of technologies that allow the information on Web pages to be tagged based on their meaning. This emerging technology would be perfect for SEAMail in that it would overlay interpretable structure on what is currently just plain text that only humans can understand.
Questions to test
For now, the next step is the CS department pilot. Genesereth will be looking not only for data on how often people use the system and how much they like it, but also to see whether SEAMail has some unintended consequences. Since the IEEE journal paper ran, for example, commentators in the media, the blogosphere, and the research community have speculated about SEAMail’s potential misuse.
“There are some significant dangers here about security, privacy and SPAM that we need to worry about before we try to roll this out to a wider audience,” Genesereth says. “That’s why we’re doing this release in stages, so that we get those things right.”
Genesereth is hopeful that the technology will prove far more useful than problematic. For instance, he points out, SEAMail exposes no data to users that they don’t already have the privilege to see. There are no plans, for example, to feed SEAMail data about faculty salaries or student grades. Database administrators and other thoughtful people still govern what data SEAMail can draw upon. It is otherwise possible, and not that invasive, to find out who has offices on the first floor of Gates, for example.
Meanwhile, Genesereth says, a semantic treatment of e-mail could filter spam even as it perhaps enables it. Because a semantic system “understands” various facts about the people associated with each e-mail address, it offers a way to filter incoming e-mail. For example, while it will be possible to use SEAMail to send an e-mail to all computer science professors at Stanford, it would also be possible to block all e-mail that isn’t from a computer science professor.
“It works both ways,” Genesereth says. “You wouldn’t get unwanted mail if you put that filter on. It’s not much worse than is the spam problem right now and it could help.”
If the computer science department test is successful, Genesereth hopes the next step will be a rollout to the entire university. From there, it could keep spreading so that someday you could address e-mails without ever addressing them at all.
