Long time coming

Ah, the curse of the rarely updated blog. The words “I’ve been meaning to write something on here for ages…” probably echo around vast chunks of the web, but it doesn’t excuse my lapse. However, I have been busy and have lots of exciting things to relate!

In short, I moved from Barcelona to London in 2014 to take up the role of Lead Data Scientist at the Future Cities Catapult. I lead the Data Science element of the Lab team, and directed our research and development on urban data, modelling, analytics and how we use all of these to make our cities better.

As of July 2015, I’ve become the Head of Data Science of an continually growing team of scientists, engineers and specialists. Fancy joining the team? Check out our open positions here.

Future Cities Catapult Logo

ACM Multimedia 2012 GeoMM Keynote

At the ACM Multimedia 2012 conference in Nara, Japan, I gave a keynote talk at the GeoMM workshop on The State of the Geotag. It is the first talk I’ve given whilst wearing socks on stage, due to the rather special venue:

That’s me on a traditional Noh stage! Thanks to workshop organiser  Martha Larson for the photo.

I’ve had requests to add a transcript or narration track to the slides, particularly as I think some of the slides may seem a bit random by themselves, so I’ll have a look into it. Seems you can ‘slidecast’ using SlideShare, which looks like it should do what I want.

I was also invited to take part in a panel discussion along with Hui Cheng (SRI international Sarnoff, USA) and Shawn Newsam (University of California, Merced), two experts in the field who had great perspectives on the challenges ahead for geo multimedia.


Placing Task MediaEval 2012

I’m now back from a very successful MediaEval 2012 held in Pisa, Italy. Pascal and I presented the chaired the Placing Task session and got to hear more about the approaches from the participating teams.

They were given a realistic sample of videos uploaded to Flickr and had to accurately annotate them with latitude and longitude coordinates using the visual video content, any available textual metadata and/or any available social data. This year, one approach was able to accurately guess the location to within 1km of the true location (which could be anywhere on the planet) 27.9% of the time on this year’s test data.


SIGIR 2012

I’m just back from giving a talk on our paper “Mining the Web for Points of Interest” (co-authors: Vanessa Murdock, Adrian Popescu and Hugues Bouchard) at SIGIR 2012 in Portland, Oregon. My first SIGIR, an interesting mix of hard-core IR theory and application, and it was great meeting some old friends.

Any questions, get in contact using the methods on the Contact page.

[nokia-maps template=”nokia.blue.place” sizes=”{‘width’: ‘auto’,’height’: ‘370’}” display_options=”no-actions no-contact” href=”http://places.nlp.nokia.com/places/v1/places/840c20fb-445b095909cc44f892394da87516c316;context=Zmxvdy1pZD1hY2QzZmRjOC1iMjI3LTQyODEtOGIzMC1hYTU4MDQyZjRlZGNfMTM0NTgxNzQwNzc5OF8wXzkxMTAmcmFuaz0z” zoomLevel=”16″ tileType=”map” title=”SIGIR 2012 at the Marriott-Downtown Waterfront Hotel”]