Wednesday, May 25, 2005
Piggy Bank brings the semantic web here and now
Piggy Bank is the rather strangely titled Firefox extension that tries to turn the existing web into the semantic web - that is to say that it aims to allow extraction of structured information from the (on the surface) unstructured data presentation that makes up most webpages, and then to go on and use that structured data.
The extension, a substantial 4.8MB download, is the result of a joint project of w3c, MIT, and HP. Many other open source projects have also been utilized in producing this one.
It provides a mechanism to capture information from the web, to save it as structured data, to tag it, and to search through and combine that information in many ways. In particular, it integrates the Google Maps mashup mechanism, allowing the data to be seen in its geographical context. Data can also be shared, by publishing it, perhaps to a central "semantic bank".
The collection of data either uses some in built collectors (for existing structured data such as RSS feeds or RDF data), or relies on specially written screen scrapers which can extract structured data from particular web sites.
Unfortunately the use of screen scrapers is often fragile - depite being launched just a couple of
days ago, I could not get any of the sample screen scrapers to work - the web pages they target seem to have changed in such ways that the scrapers no longer work. Instructions are given for writing new screen scrapers - they are written in either XSLT or javascript, and their job is to take (generally) HTML, and extract the structured data from it into RDF form.
All in all a very interesting project, pulling together a lot of smart concepts - I will be following this closely.
The extension, a substantial 4.8MB download, is the result of a joint project of w3c, MIT, and HP. Many other open source projects have also been utilized in producing this one.
It provides a mechanism to capture information from the web, to save it as structured data, to tag it, and to search through and combine that information in many ways. In particular, it integrates the Google Maps mashup mechanism, allowing the data to be seen in its geographical context. Data can also be shared, by publishing it, perhaps to a central "semantic bank".
The collection of data either uses some in built collectors (for existing structured data such as RSS feeds or RDF data), or relies on specially written screen scrapers which can extract structured data from particular web sites.
Unfortunately the use of screen scrapers is often fragile - depite being launched just a couple of
days ago, I could not get any of the sample screen scrapers to work - the web pages they target seem to have changed in such ways that the scrapers no longer work. Instructions are given for writing new screen scrapers - they are written in either XSLT or javascript, and their job is to take (generally) HTML, and extract the structured data from it into RDF form.
All in all a very interesting project, pulling together a lot of smart concepts - I will be following this closely.
Monday, May 23, 2005
Drag and drop Google results
To showcase their new AJAX like toolkit, Backbase have put together some demo sites making use of their technology.
One such demo, BBGoogle, uses the Google API to get search results, that can then be manipulated via drag and drop - at the moment the order of the results can be changed, and you can drag results out into a bookmark area. You can also drag the three "windows" around within the page, and whilst the bookmark window grows as necessary to accomodate all your bookmarks, the search results window is of fixed size - a poor design, but I guess that wasn't what they were trying to show with the demo. Overall an interesting demo, but not all that useful as it is.
Unlike the Google drag and drop implementation used by Personalized Google which updates the layout as you drag without needing you to drop it, this implementation only does the action when you drop, which doesn't quite give enough feedback for my liking in this case. Its worth noting that Google is already on (at least) it's second attempt at producing a drag and drop personalization UI - there's also one used in personalization of the Google News results. The News version does not move the item until you release it, instead providing feedback by placing a coloured rectangle where the moved item will end up.
One such demo, BBGoogle, uses the Google API to get search results, that can then be manipulated via drag and drop - at the moment the order of the results can be changed, and you can drag results out into a bookmark area. You can also drag the three "windows" around within the page, and whilst the bookmark window grows as necessary to accomodate all your bookmarks, the search results window is of fixed size - a poor design, but I guess that wasn't what they were trying to show with the demo. Overall an interesting demo, but not all that useful as it is.
Unlike the Google drag and drop implementation used by Personalized Google which updates the layout as you drag without needing you to drop it, this implementation only does the action when you drop, which doesn't quite give enough feedback for my liking in this case. Its worth noting that Google is already on (at least) it's second attempt at producing a drag and drop personalization UI - there's also one used in personalization of the Google News results. The News version does not move the item until you release it, instead providing feedback by placing a coloured rectangle where the moved item will end up.
Monday, May 16, 2005
Flickr is now reading IPTC information from images
Flickr has announced that it is reading IPTC information from images - in fact this was announced before the news about Flash, but only on the Flickr news feed, (which rather stupidly requires a login) rather than on the Flickr Blog (which anyone can read).
The discussion page on the feature documents the rules they are using as:
The discussion page on the feature documents the rules they are using as:
- IPTC keywords, when available, are added as tags.
- If there is no title specified (i.e., we were going to fall back on the image filename) and there is a IPTC *headline*, then we use that.
- If there is no description provided with the upload, then we use the IPTC *caption* when available.
- If there is no EXIF date, then use the IPTC *created date* when available.
- Add the contents of *city*, *province state* and *country* as tags when available.
"we are IPTC neophytes and we probably looked at an old spec. We will review the spec you pointed out and review for changes to make. Do you have any partcular recommendations for mappings?"so things could well change as the feature is worked upon.
Saturday, May 14, 2005
Flickr ditches Flash for photo display
Flickr is dropping Flash in favour of Javascript (DHTML, AJAX - call it what you will) for the simple act of displaying annnotated photo.
This is great news, fixing a major usability problem that plagued Flickr - obviously the level of complaints had got so loud they could not be ignored.
Update: Changed to clarify that is just the photo display function that is dropping Flash entirely. Flash will unfortunately continue to be used on other pages. There is detailed coverage of the change at Ajax Summit: Eric Costello of Flickr Presentation which gives the story direct from Eric Costello, a Flickr UI developer.
This is great news, fixing a major usability problem that plagued Flickr - obviously the level of complaints had got so loud they could not be ignored.
Update: Changed to clarify that is just the photo display function that is dropping Flash entirely. Flash will unfortunately continue to be used on other pages. There is detailed coverage of the change at Ajax Summit: Eric Costello of Flickr Presentation which gives the story direct from Eric Costello, a Flickr UI developer.
Labels: flickr
More Google Acquisitions
I produced a list before that listed all the companies that Google now owns. It's time for an update to that. The extra entries are:
- dodgeball.com - mobile phone community
- Urchin - website analytics
- zipdash - maps and traffic info for mobile devices
- Where2 LLC - Australian developers who produced Google Maps
