I think it is pretty much impossible to parse location data from stories -- 'Eastenders protestors move North from Southampton to Westminster' is going to confuse the best of systems. The only realistic way to go it for stories to be tagged as they are written.
It's not much practial help, and I don't know much about the CMS the beeb uses, but perhaps it could be done in a journo friendly way with a scenario like this: - when adding a story, they have a zoomable map (much like google!) - if a story is about a very specific point (e.g. train crash, shooting) they can pick a specific point - I'd imagine for the majority of news, rather than geo-tagging a point, they could draw one or more polygons on the map, covering the area (as large or small as necessary) that it is relevant to. So, for example, 'Blair announces in Commons extra funding for Yorkshire hospitals' the journalist could draw a polygon around the region, and a smaller one around Westminster. If there was more than one area per story, these could perhaps be ranked in importance - e.g. 1 for a polygon around Yorkshire, and 2 for the smaller polygon around Westminster. Presumably there would be a library of common areas they could select for ease of use (e.g. Westminster, each region, major cities, major motorways, etc) - I don't know if there is an existing schema for transferring polygon data (lat/long) as XML, but it wouldn't be hard to implement. This is then published in the RSS. - For the client application, it then would be simple to either plot the polygons on a map, to compare the points of the polygon with any other set of co-ordinates to see if they overlap (e.g. for the area of a Google map that is currently visible). > -----Original Message----- > From: Murray Walker [mailto:[EMAIL PROTECTED] > Sent: 09 June 2005 3:05 pm > To: [email protected] > Subject: RE: [backstage] Yet more Google Maps... > > > -----Original Message----- > > From: Dom Ramsey > > > > I was also wondering what are the chances of BBC news feeds being > > geotagged (where appropriate) > > 'fraid I cannot offer anything much here, but I can share a > tiny bit of experience fwiw. > > Some of the regional sites do have local news > http://www.bbc.co.uk/cumbria/news/index.shtml > Which you should be able to pick up as rss, see > http://news.bbc.co.uk/rss/feeds.opml > But in terms of geo-tagging, it's a bit, urm, high level. > > We've played with searching the text for known place names, > but it didn't really produce anything meaningful. We've > briefly looked at tagging articles with either one or more > x/y coord(s), or postcode(s) (or districts/sectors). I'm > sure others have done more serious study on the subject - for > me, it was mostly playing in spare/own time to see what might > be achievable. > > The main hurdle I found with the latter, aside from the work > actually tagging appropriate articles by journo/editor, is > when you have something about the "East Coast Mainline Rail > Upgrade". Which areas is it appropriate for? London, > Glasgow, every town in between, X miles either side of the > actual train line? All of the above? Does it vary by how > it's used, and if so how do you capture that meaningfully? > > Speaking very much personally, and not on behalf of gracious > employer or as part of my job, I would love to see > prototype(s) that demonstrated > > a) how a given article could be easily, meaningfully and cost > effectively geo tagged > b) how you would then use that in a productive way, beyond > what a free text search would achieve. How do nationally > relevent articles show up, without swamping the system? > > Strikes me that in a demo context, manually tagging 100 > articles need not be beyond possible, esp. if (a) gets solved. > > I know there are lots of ideas. I've not (personally) really > seen one that could be implemented within a realistic > cost/benefit kinda way, but I'm sure someone out there can fix that ;) > > </0.02p> > > > >

