Active Messenger: Email Filtering and Mobile Delivery
Download 0.67 Mb.
|
- Navigate this page:
- 4.5Wireless Application Protocol (WAP)
- 4.6Thread recognition
- 4.7Media conversion
- 4.8Content understanding, filtering, and scaling
4.4Interface to comMotionActive Messenger tries to detect the location of the user by registering the computer names where the user connects from, as well as by logging the phone numbers from where the user calls in to Phoneshell [26]. However, absolute position detection would make the location module work much more accurately. Absolute location information could be obtained, e.g., by installing active badges on radio or infrared basis. However, such a system would only work inside buildings. Outdoors, other solutions have to be found. One possibility would be to trilaterate a portable device like a cellular phone or a two-way pager. Although service providers may have this information already, it is currently not available to the user. Another option for outdoors would be the Global Positioning System GPS. This is a satellite-based radio-navigation system that is developed and operated by the U.S. Department of Defense. A project that uses GPS is comMotion [17]. It is a context-aware communication system for a mobile or wearable computing platform. A behavior-learning agent automatically learns the salient locations in the user’s life. Once these places have been determined, location specific to-do lists, electronic Post-it™ notes, or reminders and subscription requested information can be delivered to the user at the right place and time. The user interacts with the system through a visual and/or speech interface. Because comMotion already knows the users location, Active Messenger could query this information and enhance with it the accuracy of the existing location detection. 4.5Wireless Application Protocol (WAP)WAP44 is a joint effort between several key players of the mobile communication industry to further develop the ideas of intelligent messaging and other similar technologies. The goal is to develop an open protocol that can be adopted by any manufacturer and that enables easy delivery of relevant information and services to mobile users. WAP is HTML-compatible and supports GSM networks and CDMA. However, the aim is to support all other current and future digital cellular technologies. WAP allows users of cellular phones supporting the protocol access to applications and functions such as:
Because Active Messenger addresses similar problems, it seems to be appropriate to keep in touch with this development, possibly integrating WAP capable devices once they will be available. 4.6Thread recognitionQuite often, email communication between users can be clustered into threads. This means, a user refers to one or several prior messages concerning the same issue. It would be interesting to have the agent detect a thread and the messages that are part of this thread. If Active Messenger could detect a thread reliably, messages of the same thread could be grouped together and treated in a special way. However, thread recognition is not trivial. A simple way would be to group messages that come from the same sender and have the same subject line. This approach works only if the message from a person is always a reply to the last message sent to this person. If user A sends two messages to user B, and user B replies to one of them, it is only possible to detect a thread if the subject line is left relatively unchanged. Additionally, recognizing threads that span more than one medium is difficult because it may require content understanding. E.g., if the user sends email to a person, and this person calls the user back and leaves a voice mail message, it requires reliable speech recognition and content understanding to group these two messages together. 4.7Media conversionActive Messenger is capable of converting email messages to another medium, e.g., email to fax, or email to voice mail. The agent performs simple conversions, e.g., it modifies email messages that have to be sent to a pager by deleting signatures and limiting the length of the messages. Furthermore, email messages that have to be read to the user on the phone or sent to the voice pager are processed by Phoneshell [26] to enhance the intelligibility of the synthesized audio. However, it would be interesting to modify the agent so that the device characteristics are taken into account more extensively when converting messages. The output capabilities of each device and channel are different. It would be interesting to have an agent that is able to determine the nature of the device the user is using, and deliver content in a format most suitable for its output capabilities. E.g., if the user receives a message that contains a picture, the agent decides if the picture should be sent to the device separately, depending on the capabilities of the device to display graphics, perhaps adjusting the picture resolution and size according to the available bandwidth. Similarly, sound files embedded in email message could be transmitted as audio when the agent calls up the user on the phone. 4.8Content understanding, filtering, and scalingSimilar to media conversion, but much harder to realize, would be content understanding. This means to modify the agent so that it is able to understand the actual content of a message. This would be useful to scale up or down the message content, or adjust the forwarding rules of the agent. For example, if a device can only display short messages, the agent would summarize a long email message so that it fits the display without losing the essential information. This is only possible if the agent understands the content of a message. On the other hand, if a short message contains abbreviations, Web links, or people’s names, the agent could enhance the message with information from the Internet and other resources. Furthermore, if the email contains a table of data, e.g., stock quotes, and the user’s device has a graphic display, the agent could transform the table into a graphic and send this visual information instead. Content understanding would also facilitate thread clustering. Additionally, if the agent would be able to understand the content of a message, it could determine the importance of a message more accurately. However, it is clear that content understanding in general is a difficult problem. Obviously, such a system would require extensive use of artificial intelligence and is therefore far from being realized soon. Directory: ~stefanm ~stefanm -> Mit media Lab Stefan Marti mit media Lab running head: Active Messenger Download 0.67 Mb. Share with your friends:
|