Google Translate is now instant. But still fun (and dangerous).

Google Translate has already accelerated my Welsh learning. It helps to decipher a daunting piece of text.

Now Google Translate is instant. They changed the interface slightly and it flashes up the equivalent translation as you type. Boy.

In other words you get the same flawed “translations”, now even faster!

Try it for Welsh to English.

Example phrases:
Dw i’n cyfieithu.
Defnyddia yn ofalus.
Gwlad beirdd a chantorion, enwogion o fri

I wish there were a proper online Welsh-to-English dictionary that did instant look-ups. It would take some of the friction out of reading difficult books. Just leave the laptop open, type a difficult word and get the meaning NOW.

Having to click is too slow a method because it breaks the flow of the book. Reaching for a dictionary is even worse. The look-up needs to be as near to the speed of thought as possible.
🙂

I say “proper dictionary” because Google still gets words wrong y’see. It’s based on statistical translation and uses the “most likely” translation based on a corpus of text equivalents in both languages. It also seems to have a limited vocabulary.

And a reminder…

Don’t use it for roadsigns! But you can use it to check the gist of a professional’s translation…

Sock And Awe Google Analytics (Just A Flash In The Pan?)

sock and awe

After the Bush/Shoe incident, anyone who’d spent even a few moments in eccentric corners of web knew there would be a creative response online. And it came. Wired has a summary of the shoe-inspired games and animations.

So Sock And Awe wasn’t the only Flash game based on the Bush/Shoe event. But it was the best.

Now Alex Tew, its creator, has sold the site as a property on eBay for £5,215. Whichever way you look at it, that’s a good rate for a few hours of work – not to mention the email subscribers he gathered, which were not part of the sale.

Rory Cellan-Jones at BBC News has the details of this high speed micro-start-up.

As Cellan-Jones notes, the site is based on a current affairs event and will now rapidly decline in value. It’s up to the new owners to extract value from it.

But this doesn’t detract from the cheek and verve of Tew and his colleagues. Everything from the choice of name to the design to the speed of launch and then the one-day auction was executed with skill. See also: his Million Dollar Homepage. If you’re curious about his next move in the world of the web, check out Tew’s forthcoming start-up PopJam.

While the Sock And Awe site was being auctioned, I contacted Tew and asked to see the full visitor stats, via Google Analytics. The visitor counts and top countries were already generally known, but I wanted to see precisely what was happening.

Now that the sale has closed and people are chatting about it, the full analytics make interesting reading so I thought I’d post them up here – with first some graphical highlights then the unexpurgated PDF dumps. (“After the jump”, if you will.)

It’s a good case study in site design and branding.


Sock And Awe – Visitors Overview
The bounce rate is high, which for the average site would normally be very bad. (In other words, most people are just looking at the homepage then leaving.) But Shock And Awe is mainly about the homepage, so it’s an exception to most sites.


Sock And Awe – Top 20 Referring Sites (Detail)
Most visits are getting there by typing into the address bar. Far fewer are clicking to come from other sites. This shows the value of having a good web address that’s memorable and easy to spell. Notice I said “visits” rather than “visitors” (uniques). As you can see from the New vs. Returning PDF, 13% of them are repeat visitors, presumably returning to play again.


Sock And Awe – Map of Visitors
An unusual sign of accord between USA and France, who occupy the top two spots. Google Analytics also records “not set” for country unknown but this is much further down the chart at position 42. Middle Eastern countries can’t get enough of the Bush bashing, as you can see from the full countries PDF

All of the analytics in my blog post here were taken at around 6:15AM GMT on Thursday 18th December. As you can see, the graphs and figures plummet on the 18th because they’re not showing a full day’s stats. It may be better to disregard that day’s totals and regard all analytics as a snapshot showing qualitative insights.

Grab the ZIP file of all sockandawe.com analytics. Or view individual pages below.

Dashboard
This is the overview data.
Dashboard

Visitors
Check out Time on Site for All Visitors – the earlier visitors have much longer attention spans!
Visitors Overview
Map Overlay
New vs. Returning
Languages
Visits
Page Views
Absolute Unique Visitors
Bounce Rate
Time on Site for All Visitors

Traffic Sources
As I mentioned before, the direct traffic is by far the largest. With no time to mount an SEO campaign, Sock And Awe still captures some keyword search traffic, again thanks to the memorable name. (Google and other engines recognise matches with the domain name.) It also captures a few who mistakenly type the URL into their search bar instead of their address bar. (Incidentally, you may be wondering why my own personal blog is called, of all things, Quixotic Quisling. Well, I like to play the long game.)
Traffic Sources Overview
All Traffic Sources
Direct Traffic
Referring Sites
Search Engines
Search Keywords

Content
Content analytics are perhaps the least interesting because this site has very few pages. Although Top Content does give a hint how many people attempted to sign up for the newsletter – at least 30,000 it would appear. (After signing up, they arrived at sockandawe.com/email for a confirmation message. This folder has now been removed.) I say “attempted to sign up” because my own experience is that many people try search queries in these boxes, even despite clear labelling. Tens of thousands is still a good order of magnitude, even if half are bad. Many of the web addresses listed in Content account for framed visits (one recognisable example is somebody using Babelfish, in vain, to attempt to get a translation).
Content Report
Top Content
Top Landing Pages
Top Exit Pages
Average Pageviews

Quick word on Development Costs
According to reports, the game took a night to build. The game engine is very simple – if you think about it, it just compares the X-position of the mouse cursor (which is invisible) with the random X-position of Bush’s head graphic. If the distance is within a pre-set striking tolerance, then the whack graphic is shown. I would say the most time consuming part of the Flash game development was designing the graphics and animation.

Similarly, the bandwidth costs would be low. If you run the site through an analyser, it’s currently around 200kb of data. The site has been slightly modified to remove the subscription option and add advertising, but these are not big changes.

Discussion
Can you glean any more significant insights from these stats? You can comment or send me a message on Twitter.

The Freaky Future of TinEye

In a dizzying round-up, Reportr alerts us to 10 tech trends, the 11th presumably being to drop letter ‘e’s now that single-word domains are in short supply (cf. Flickr, Dopplr, Tumblr…)

The thing that caught my attention was the mention of visual search. I’ve been playing with a new search engine, TinEye, for a couple of months. It claims to be the first on the web – it’s probably the best example so far.

You submit an image to TinEye and it returns similar images from all around the web, based on pattern matching.

Their page of cool TinEye searches is a good introduction.

This is a nothing shy of a REVELATION. It’s funny how tech journalists remain all cool when they see something like that. Well I’m going to point out it’s AMAZING. Inevitable maybe. Given there is so much information on the web, there do remain undiscovered and undeveloped ways in which we can retrieve it. Visual search is a known problem – it just required somebody to sniff out the winning search index algorithm, for accuracy and speed.

In the life of the web, we have become accustomed to fairly good text-based search. The search box has become a second brain for many. We live in a post-search world. (At least, those of us who live in countries with easy access to the web do.)

How does Google find the stuff we want? At the moment a particular page or file has a URL. It may have other metadata such as a filename and keyword tags or alt tags. And unless you’re talking about a video or image, it could be a document with some kind of text body. It also has the pages from around the web which link to it. Those are the only clues, unless I’m forgetting any.

It’s pretty easy for computers to do pattern matching on text. Good search algorithms that work on text databases have been known for decades. But it’s the indexing of the whole web that’s always the big challenge. When you do a Google search, you’re not really searching the web itself, you’re searching Google’s indexes.

What is so amazing about TinEye is they are introducing another clue into the search – the content of the image – and found a quick enough way to index it.

Before I get too excited about TinEye, right now you can hope for hit-and-miss results at best, particularly as the pool of indexed images is relatively low. Since some early announcements of the service this has increased and they recently widened this pool to 901 million images, some of which might be yours.

It remains at the beta test stage, but it’s already quite useful. Some photographers and other visual artists have already been able to track down where their images have been used.

Visual search is still in its infancy and we can expect other players to possibly rival TinEye. There’s a shopping service called like.com mentioned in the Reportr trends piece, but I can’t seem to get past the nauseous feeling of entering a glitzy shopping mall the minute I arrive. It does pose another question for me.

What about the commercial outcomes of visual search? How will companies try to seek traffic from visual search? Will there be attempts at search engine optimisation? Will we see images being priced and ads being sold like Google AdWords? (GOOG)

TinEye doesn’t do human face recognition yet. Could it be just a matter of time, processing power and a dash of ingenuity? Let’s imagine a world where it is possible and it becomes a commonplace thing for use and abuse. Permit me to do some wild speculation.

What about finding doppelgangers and lost relatives? Missing people? It’s possible that a missing person could be an incidental feature of a photograph that somebody posts online.

Could it be used for casting films or theatre, when a new Kubrick absolutely has to find an actor with distinctive facial features?

What about looking for a partner? In a world already filled with all manner of weirdness, would someone try to seek a replacement partner who looks like their ex?

How could police use and potentially misuse this technology? Could they find missing suspects? Match fingerprints?

The future of TinEye and visual search could be like having eyes everywhere… It’s very promising and possibly a bit unsettling.