JANUARY/FEBRUARY 2009 49
Semantic Email Addressing
cally dened set o entities (see the “Related Work in Semantic Email Addressing” sidebar or other research in this area). A SEA mailserver computes the recipients o a semanti-cally addressed email on the fy based on theaddress’s semantic denition. SEA has other benets as well. First, it doesn’t require dis-covery. In traditional email systems, it’s di-cult or humans to discover email addressesand mailing lists, and automatic discovery bycomputers is almost impossible. In addition,SEA requires no maintenance. In traditionalmailing lists, an administrator must create andmaintain the list and individuals subscribe or unsubscribe to the list, and change their inor-mation with regard to the list. This can be par-ticularly onerous when users must deal withmany lists — or example, when their emailaddresses change. Instead, users can updatetheir personal inormation, such as email ad-dress, once, in a single place, and the SEA mailserver would automatically adapt.SEA has application in both corporate in-tranets and the Internet. It also raises severalissues, including security and privacy issues,errors, user adoption, and standardization.
To illustrate SEA, we consider two examples.
Corporations and other organizations oten havedatabases o inormation about personnel, proj-ects, customers, and so on. Users can leveragethis inormation to send emails based on prop-erties o the people in a particular department.Consider a moderate-sized company withseveral departments. The company has a cen-tralized database containing the ollowing in-ormation about its personnel: name, emailaddress, department, group, position, project,and date o hire. Given this inormation, a user might dene groups o people by querying thedata — or example, or “all senior managers inthe accounting department.”The company might also have the ollowinginormation about its projects: project name, pri-ority, leader, start date, and end date. With this
Related Work in Semantic Email Addressing
everal other researchers have recognized the value in bring-ing semantics to email. For example, the Inormation Lenssystem
lets users send semistructured email messages and l-ter those messages using production rules. Users can send toa special mailbox called “anyone,” and anyone can choose toreceive messages rom this mailbox based on production rules.This fips the nature o widely broadcast emails on its head.Instead o starting with receiving all emails and whittling themdown based on ltering rules, the user starts with an emptyinbox and pulls in email o interest. This is similar to the RSSsubscription model. As RSS eeds contain more semantic inor-mation, the semantic subscription model exemplied by Inor-mation Lens might become more commonplace.More recently, MailsMore
lets users annotate an email’scontent with Resource Description Framework (RDF) triplesand automatically includes RDF triples based on standard emailheaders such as the “To,” “From,” “Subject,” and body elds.This can be used or semantic ltering and ling o emails.The Mangrove system
takes this idea urther. It allowsnot only structured email content but also semantic email proc-esses. Users can script email clients with declarative workfowsthat automatically aggregate inormation obtained rom manyemail responses, automatically resend emails to people whohaven’t responded, or analyze the semantic content o incom-ing email messages and respond accordingly.Most relevantly, Microsot Exchange 2003 lets administra-tors create query-based distribution groups, which are essen-tially mailing lists whose recipients are based on a LightweightDirectory Access Protocol (LDAP) query run when the email issent. This alleviates much o the administrative work requiredto maintain a mailing list. However, because only an administra-tor can create the mailing lists, users can’t send SEA mail, andthe inormation upon which the lists are based isn’t under us-ers’ control. None o this application’s unctionality is availableto users and very little to administrators. In act, users can’tsee that a distribution group is query based: each query-baseddistribution group has a name, so to an outsider looks like aregular mailing list.
T. Malone et al., “Intelligent Inormation-Sharing Systems,”1.
, vol.30, no. 5, 1987, pp. 390–402.A. Kalyanpur et al.,2.
Representation Formalisms and Methods,
tech. report,Univ. o Maryland, 2001; www.mindswap.org/papers/SMORE.pd.O. Etzioni et al., “Semantic Email: Adding Lightweight Data Manipulation3.Capabilities to the Email Habitat,”
Proc. 6th Int’l Workshop the Web and Data-bases
, ACM Press, 2003, pp. 12–13.L. Mcdowell, O. Etzioni, and A. Halevy, “Semantic Email: Theory and Appli-4.cations,”
Web Semantics: Science, Services and Agents on the World Wide Web
,vol. 2, no. 2, 2004, pp. 153–183.L. McDowell et al., “Semantic Email,”5.
Proc. 13th Int’l Con. World Wide Web
(WWW 04), ACM Press, 2004, pp. 244–254.
Authorized licensed use limited to: IEEE Xplore. Downloaded on April 22, 2009 at 08:07 from IEEE Xplore. Restrictions apply.