Re: [backstage] Geotagging BBC news stories
I've been thinking about servers wrt annotations. Why wouldn't you just have a server that takes a query with a guid and produces an annotation? I thin kthere is a really good reason for not going for this at the first cut: simplicity. If everyone out htere who is providing meta data for rss feeds has to also provide a server, then that is a high barrier to entry. If all they have to do is run a web server and ping a ping service, that is easier. It would then be trivial to create a server that reads these annotations and acts as a clearing house with as much query based service as possible. Peter, your comments about location are spot on. In fact I tag all my rss feeds, where possible with locatoin data (where the feeds are published). I know that serious feed based services (like moreover.com) tag feeds with a huge number of meta data (40?). Having a server that, given a location, produces coords - this makes sense for fixed meta data, but then, if this is fixed, it would be preferable to only look the infor up once when entering the feed into a database, not to do it every time an article from the feed is published. I think one of the things you are getting it, which sounds really cool, is something that would be able to say 'here is how people in turkey view the bombings in london'. Perhaps you are extending the annotation idea (which is item oriented) to feeds. In other words, annotations for a feed (including the information about where the feed is from) could be delivered by a lookup service? Matt On 7/23/05, Peter Williams <[EMAIL PROTECTED]> wrote: > Hi > > a couple of thoughts > > The use of Geotagging could be more useful if the tags were split according > to the location of the author, location of the central story elements and > the location where the story was published. For example, a story about > Mugabe (location tag Zimbabwae), is filed from South Africa (location Tag > Joberg), and "published" in the UK (location tag London/UK) > > In this way it would be possible to group stories by perspective, i,e the > Australian press version of the London bombings, or the UK perspective on > Mugabe as well as grouping by location of the stories, and their authors. > This may link in well with any catagorization approaches to news feeds. > > A look up server that contained well known locations expressed in Lat/Long > would make the application of the tag a simple process. If extended to the > annotators, then one could the see the location based perspective of the > commentator. > > > Regards > > > Peter > > (located in Australia- somewhere in the south pacific) > > > > > > > Matthew Hurst wrote: > More thoughts: the annotation spec should include information about when > the > rss feed was crawled and also when the annotation stream was created. > > I have some data now (both your annotation and some rss files with > matching guids) saved to disc, I'll post the results sometime this > weekend. > > MattH > > On 7/22/05, Matthew Hurst <[EMAIL PROTECTED]> wrote: > > > This is great. I will get something together asap. > > Thanks! > > Matt > > On 7/22/05, Davy Mitchell <[EMAIL PROTECTED]> wrote: > > > On 7/22/05, Matthew Hurst <[EMAIL PROTECTED]> wrote: > > > I'm still for something like the original file format I suggested > earlier. Simplicity is very attractive in this type of endevour. > > :-) > > http://www.latedecember.com/sites/moodnews/moodnewsanno.xml > > Its one big xml of all the feeds as I don't store the feed URL in the > db (yet!). Should be enough for a demo... Email me if you need > anything changed etc. > > Interesting stuff!! > > Mood News development trundles on with a mix of back end and client > stuff being added. It has been slowed by me hunting for the perfect > Python IDE. A new version next week perhaps? > > Thanks, > > Davy Mitchell > > http://www.latedecember.com > > > - > Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, please > visit > http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html. > > > > - > Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, please > visit > http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html. > > > > > -- > Peter Williams > CEO > Eurofin Technologies > > www.eurofin.net > > T +612 9968 1240 > F +612 9968 1552 > - Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, please visit http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html.
Re: [backstage] Geotagging BBC news stories
Hi a couple of thoughts The use of Geotagging could be more useful if the tags were split according to the location of the author, location of the central story elements and the location where the story was published. For example, a story about Mugabe (location tag Zimbabwae), is filed from South Africa (location Tag Joberg), and "published" in the UK (location tag London/UK) In this way it would be possible to group stories by perspective, i,e the Australian press version of the London bombings, or the UK perspective on Mugabe as well as grouping by location of the stories, and their authors. This may link in well with any catagorization approaches to news feeds. A look up server that contained well known locations expressed in Lat/Long would make the application of the tag a simple process. If extended to the annotators, then one could the see the location based perspective of the commentator. Regards Peter (located in Australia- somewhere in the south pacific) Matthew Hurst wrote: More thoughts: the annotation spec should include information about when the rss feed was crawled and also when the annotation stream was created. I have some data now (both your annotation and some rss files with matching guids) saved to disc, I'll post the results sometime this weekend. MattH On 7/22/05, Matthew Hurst <[EMAIL PROTECTED]> wrote: This is great. I will get something together asap. Thanks! Matt On 7/22/05, Davy Mitchell <[EMAIL PROTECTED]> wrote: On 7/22/05, Matthew Hurst <[EMAIL PROTECTED]> wrote: I'm still for something like the original file format I suggested earlier. Simplicity is very attractive in this type of endevour. :-) http://www.latedecember.com/sites/moodnews/moodnewsanno.xml Its one big xml of all the feeds as I don't store the feed URL in the db (yet!). Should be enough for a demo... Email me if you need anything changed etc. Interesting stuff!! Mood News development trundles on with a mix of back end and client stuff being added. It has been slowed by me hunting for the perfect Python IDE. A new version next week perhaps? Thanks, Davy Mitchell http://www.latedecember.com - Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, please visit http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html. - Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, please visit http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html. -- Peter Williams CEO Eurofin Technologies www.eurofin.net T +612 9968 1240 F +612 9968 1552
Re: [backstage] Geotagging BBC news stories
Davy - yes, I think there is a real opportunity here for something. I'm imagining a ping server for FAS data. There is open source java code for a ping server, I bet we could trivially mod it to work for FAS data. If we could put this up and provide a front end (very very simple thing, just a list of all the annotation streams that have updated in the last n minutes)... What do you think? Do you have any server resources that we could use? If not, I'm pretty sure I can get my hands on something. BTW, note that the annotationURI i the annotationStream tag is meant to point to a specification for the annotations in the node, not a pointer to the actual file. In otherwords, you need to have a mood annotation specification somewhere (e.g. stating that negative means bad, positive means good and 0 means neutral). The way URIs work is that they don't have to actually resolve to something, they just have to *mean* something. However, it is far more convenient if the URI points to an actual document that states the spec. I will work on something more formal as a definition. MattH On 7/23/05, Davy Mitchell <[EMAIL PROTECTED]> wrote: > On 7/23/05, Matthew Hurst <[EMAIL PROTECTED]> wrote: > > Here is the blog entry: > > > > http://datamining.typepad.com/data_mining/2005/07/feed_annotation.html > > > > Feed Annotation Streams > > Looks good - but I might have to hack the Python RSS library I am > using to get the GUID tag out of it :-) I'll have a go at updating the > XML with more data such as supplying the colour data used on the web > page instead of the raw score. > > If 2 or 3 of these feeds are combined. well the mind boggles! > > Off bug squishing... > > See you, > Davy > > -- > > Davy Mitchell > > http://www.latedecember.com > > > - > Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, please > visit http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html. > - Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, please visit http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html.
Re: [backstage] Geotagging BBC news stories
On 7/23/05, Matthew Hurst <[EMAIL PROTECTED]> wrote: > Here is the blog entry: > > http://datamining.typepad.com/data_mining/2005/07/feed_annotation.html > > Feed Annotation Streams Looks good - but I might have to hack the Python RSS library I am using to get the GUID tag out of it :-) I'll have a go at updating the XML with more data such as supplying the colour data used on the web page instead of the raw score. If 2 or 3 of these feeds are combined. well the mind boggles! Off bug squishing... See you, Davy -- Davy Mitchell http://www.latedecember.com - Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, please visit http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html.
Re: [backstage] Geotagging BBC news stories
Here is the blog entry: http://datamining.typepad.com/data_mining/2005/07/feed_annotation.html Feed Annotation Streams With the help of Davy Mitchell, and the input of others on the backstage discussion list (Tony Hirst, Duncan Barclay, and others) we have put together the first proof of concept of Feed Annotation Streams (formerly known as RSS Annotation Streams). The idea is simple - consumers of a feed create annotations for that feed, the put those annotations in a new feed that refers to the original items by guid. The example is based on the BBC RSS news feed and Davy's mood score (is the news good or bad?) for the articles. The RSS feed is consumed in DataSphere and geotagged, the items are then referenced against the mood score FAS and coloured. The colouring I applied is very simple. If the sum of the mood scores for a location is positive, the pin is painted green, negative red and zero white. There is a lot still to do in terms of defining the spec, but it looks like it has great potential. Fas The Feed Annotation Stream looks something like: On 7/22/05, Matthew Hurst <[EMAIL PROTECTED]> wrote: > Ok - here is the very first instance of using a Feed Annotation Stream > together with a feed from the bbc to produce geolocated, mood mapped > news. The image shows a sample of data from a recent grab of rss from > the beeb mixed with reading Davy's mood annotation > feed. The pins indicate the volume of posts and the colours indicate > the sum of the mood scores for that locatoin was negative (red) and > positive (green). > > Phew! > > Having done this (with throw away code) I have some comments: > > 1) guid - needs to be the guid from the rss feed, not the (perma) link > to the story. > Davy - I munged the urls you had in there to provide actual guids. > 2) unique? there were some duplicates, but i guess this doesn't matter. > 3) Feed Annotation Stream is a better name as there is no reason it > has to be for > rss - anything with an identifiable guid would do. > > I'm probably going to blog this later and repeat the post to this > list. I'd also like to > put up a complimentary FAS for this data with the geolocation stuff in > it to give a complete > proof of concept. > > Davy - great stuff here - let me know what you would like to do next. > I am working towards getting my stuff visible outside the company - > these wheels move a little slower as there are some purchases > involved. I would like to get more work done on the spec and to host > it off my blog somehow. Please throw any comments up on this list! > > MattH > > On 7/22/05, Matthew Hurst <[EMAIL PROTECTED]> wrote: > > More thoughts: the annotation spec should include information about when the > > rss feed was crawled and also when the annotation stream was created. > > > > I have some data now (both your annotation and some rss files with > > matching guids) saved to disc, I'll post the results sometime this > > weekend. > > > > MattH > > > > On 7/22/05, Matthew Hurst <[EMAIL PROTECTED]> wrote: > > > This is great. I will get something together asap. > > > > > > Thanks! > > > > > > Matt > > > > > > On 7/22/05, Davy Mitchell <[EMAIL PROTECTED]> wrote: > > > > On 7/22/05, Matthew Hurst <[EMAIL PROTECTED]> wrote: > > > > > I'm still for something like the original file format I suggested > > > > > earlier. Simplicity is very attractive in this type of endevour. > > > > > > > > :-) > > > > > > > > http://www.latedecember.com/sites/moodnews/moodnewsanno.xml > > > > > > > > Its one big xml of all the feeds as I don't store the feed URL in the > > > > db (yet!). Should be enough for a demo... Email me if you need > > > > anything changed etc. > > > > > > > > Interesting stuff!! > > > > > > > > Mood News development trundles on with a mix of back end and client > > > > stuff being added. It has been slowed by me hunting for the perfect > > > > Python IDE. A new version next week perhaps? > > > > > > > > Thanks, > > > > > > > > Davy Mitchell > > > > > > > > http://www.latedecember.com > > > > > > > > > > > > - > > > > Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, > > > > please visit > > > > http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html. > > > > > > > > > > > > - Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, please visit http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html.