The
way we navigate in cities has been revolutionized in the last few years by the
advent of GPS mapping programs. Enter your start and end location and these
will give you the shortest route from A to B.
That’s
usually the best bet when driving, but walking is a different matter. Often,
pedestrians want the quietest route or the most beautiful but if they turn to a
mapping application, they’ll get little help.
That
could change now thanks to the work of Daniele Quercia at Yahoo Labs in
Barcelona, Spain, and a couple of pals. These guys have worked out how to
measure the “beauty” of specific locations within cities and then designed an
algorithm that automatically chooses a route between two locations in a way
that maximizes the beauty along it. “The goal of this work is to automatically
suggest routes that are not only short but also emotionally pleasant,” they
say.
Quercia
and co begin by creating a database of images of various parts of the center of
London taken from Google Street View and Geograph, both of which have
reasonably consistent standards of images. They then crowdsourced
opinions about the beauty of each location using a website called
UrbanGems.org.
Each
visitor to UrbanGems sees two photographs and chooses the one which shows the
more beautiful location. That gives the team a crowdsourced opinion about the
beauty of each location. They then plot each of these locations and their
beauty score on a map which they use to provide directions.
The
idea here is that the user enters a start and end location and an algorithm
then finds the most
beautiful route, rather than the shortest one. It does this by searching
through every possible route, adding the beauty scores for each and choosing
the one that ranks highest.
Quercia
and co say that on average these routes turn out to be just 12 percent longer
than the shortest routes, which makes them reasonable alternatives for a
pedestrian.
To
work out whether the routes chosen by the algorithm are really more beautiful,
Quercia and co recruited 30 people who live in London and are familiar with the
area, to assess the recommended paths. And indeed, they agreed that the routes
chosen by the algorithm were more beautiful than the shortest routes.
But
that’s just the start. Crowdsourcing opinion for every possible location in a
city is clearly a time-consuming and potentially expensive business. So Quercia
and co have automated this process using photos from Flickr and the data and
tags attached to them.
They
chose some five million pictures taken in the same places as their original
photos and then mined the data associated with them to see what parameters
correlated with beauty.
Factors
that turn out to be a good indicator of beauty are things like the number of
pictures taken of a particular scene and comments associated with positive
emotions. So looking for locations on Flickr that fulfill this requirement
ought to produce a list of beautiful places in any city.
Quercia
and co tested this idea in Boston to find beautiful locations on Flickr and then
used their algorithm to find the most beautiful path between two locations.
They then asked 54 people to evaluate these paths. Sure enough, the
participants generally felt that the routes chosen by the algorithm were more
beautiful than the shortest parts.
If you
know Boston or London yourself, you can evaluate the routes chosen by the
algorithms yourself by examining the maps in the paper.
Of
course, there are potential problems. Some locations are less attractive at
certain times of the day, for example during rush hour when traffic is heavier
or at night when the character of some parts the city can change dramatically.
The algorithm cannot account for these differences
Nevertheless,
this is an interesting approach that has the potential to change the experience
people have in interacting with the city. It’s not hard to imagine that tourist
authorities might use an application like this to help visitors experience the
best parts of a city on foot.
Quercia
and co have a plan like that. Their next goal is to build a mobile app and test
it in the wild across different cities in Europe and the U.S. Keep an eye out
for it.
No comments:
Post a Comment