Web Calendar auto-generated from mailing lists

FROM: Peter Groves

What NEED does this meet?
Web calendars (specifically activist calendars) have a significant problem in that every event must be manually added to all web sites the poster wishes to advertise on. As a result it's rare that they are actually up-to-date and contain much useful info. In contrast, mailing lists for various groups have lots of up to date info for people who have specifically asked to receive updates on the activist group, sports club, band, etc, that the list is about. The information is there for people who know to ask for it. There is a chicken and the egg problem, though: how do people know to request info on your events if they don't know what your events are? On a college campus, for instance, hundreds of groups are having events all the time, from weekly meetings to speakers to protests, but often the only ones who know about them are people who are already in the group.

What is the APPROACH?
Information extraction (related to information retrieval - the broader category web search engines fall under) is the process of filling in a database with information found in text. There's quite a bit of research in this area in Universtities, but little commercial applications (one company makes software to pull out name, skills, yrs. experience, etc from resumes submitted electronically). If you could get a few volunteers to search the web for every mailing list that might announce events, you could sign up some central email address to receive the mails. Then as the emails came in, an information extraction engine could automatically populate a calendar by pulling out the date, time, location, and event title (more likely it would be a summary because the exact title is difficult to identify, but that's a technical issue to be worked out later). This could be combined with some sort of geographic information backend (some such projects have been submitted, such as "collaborative cartography system") so that people could even do geograpic based queries. If we wanted to be ambitious, we could even create personalized feeds based on where people live and what they're interested in that could be handled by some open source calendar software, and people's calendars could be updated automatically.

What are the BENEFITS to people?
Currently, people announcing events must manually submit the info to each calendar site they want to be seen on. If the sites were using this technology, the mailing-list operators could simply add them to the mailing list and the information they sent to their members would automatically get displayed. If many groups started doing this, the quality of community web calendars would increase significantly and people would be more likely to visit them, or even request an rss feed for events in their area.

What is the COMPETITION?
The competition is every web-based community calendar that requires you to fill out a form to submit an event, or requires the calendar-maintainer to do it themselves from information in an email they receive.

What BUDGET & LOGISTICS are required?
There is a decent amount of open source code that could be exploited (the data-base backend, mail management, html calendar display, etc). There is some information extraction code (like GATE from the University of Sheffield), but that's usually written in java, and this may be computation-intensive enough to warrant rewriting in a different language (like ocaml, I have no interest in dealing with C/C++ code).

If we wished to simply write/maintain the application, the only cost would be developer time. Existing calendar sites could simply use our application. Running a server wouldn't be hard, however, it would just take a machine to do the processing and another machine to be the actual web server. It would probably require someone to fix incorrect postings, which may be a full time job if thousands of emails are being handled.