Information

How well does Google.im perform?

Watching a video today about Google's localisation of their search results (part of "A peek into our search factory"), I spotted the Manx flag and thought I'd give Google's Isle of Man localised search (www.google.im) a test run and see how it performs at disambiguating queries and giving preference to local search results.

Topic

Google Isle of Man

Google UK

Places

Ballasalla

all relevant

all relevant

Castletown

top 2 relevant, rest contain Irish, Scottish and Manx results

top 2 relevant, rest contain Scottish and Manx results

Colby

X no relevant results (mostly US)

X no relevant results (UK/US mixture)

Douglas

X no relevant results

top 2 relevant

Peel

X no relevant results

2 of top 3 results relevant

Ramsey

X no relevant results

X no relevant results

Sport

Golf

X no relevant results

X no relevant results

Golf courses

X no relevant results

X no relevant results

Swimming

X no relevant results

X no relevant results

Swimming pools

X no relevant results

X no relevant results

I think that's enough testing to realise that the Google Isle of Man tailored service isn't Isle of Man tailored at all. Even the Google UK service is better.

It's great that they have a service aimed at users from the Isle of Man, and I'd love to see it succeed, but I don't understand why they have it out there at the moment if it's not actually tailoring the search results to be useful to the Manx market.

Help Google do their job better

I'm in my last week here in Germany, and very much looking forward to moving back to London to live and work, but unfortunately there are all sorts of things that need to be arranged before I move. That includes things like working out what I need to do with my bank account, pension, insurance, and other such things, all of which are really stretching my (much improved, though still very limited) knowledge of the German language.

So that means that every now and then I need to resort to dictionaries (dict.leo.org is a great community-built one) or machine translation software such as that used by Google Translate to give me a better idea of the information that I'm trying to read.

Trying Translate today on the Deutsche Rentenversicherung (German Pension Insurance) website, I noticed something that I hadn't noticed before. It may just be that I don't often translate whole web pages, but when trying that today I noticed that on each phrase you hover your mouse over, you get a little popup and the option to 'suggest a better translation', thereby giving a feedback loop to continuously improve the machine-translated text.

Suggest a better translation for Google

There goes Google again, using individuals in the huge crowd that is their user base to actively improve their information services for all, and for free. They also announced this week that they will be allowing people to tweak address locations, improving their geocoding capabilities. It's not quite at the level of their efforts in India to let people generate map data for them, but it's starting to look like that's the direction it's moving in... and it's all with the aim of improving information availability around the world, so I'm all for it.

Update: It looks like Google switched to their own in-house machine translation system last month, but has been letting people suggest new translations for some languages since earlier in the year.

Express ticketing that isn't

Sitting on a National Express coach from Heathrow to Gatwick, I'm pondering the failures of system design of some so-called express ticketing systems, namely the one National Express has rolled out to its coach stations. For my trip back home I'd bought a Young Persons Coach Card online for £10 in the knowledge that I'd save on my tickets (I don't recall why I didn't buy tickets at the same time, but I wish I had).

Walking into the Central Coach Station at Heathrow and seeing the queue for the ticket counter, the natural thought may be to try out one of the Express Ticket terminals that isn't in use. Thinking I could avoid the queue, I went and touched the screen to start the process. A screen of popular destinations is displayed. So far, so good.

The first hurdle is right there though. Which terminal at Gatwick am I flying out of? I dig into my bag for my iPod, checking its calendar to see if I noted it down. No mention of it - my fault for not thinking about that when I put the flight into my calendar - but it'd be nice if the system asked who I was flying with and pointed me in the right direction.

Taking a guess at the South Terminal, I then wonder what the difference is between a single and a return. Do I get any savings for buying a return? Does it just mean I can do it in one go without having to hassle with ticket machines on my return? No assistance there, so already a little frustrated, I select a return ticket.

Choosing the first coach leg was easy, I just chose one of the ones in 20 minutes or so. Does the specific coach really matter? What happens if I've not finished this process and got to the bus by that point? Can I travel on a different one instead?

Now for the pain of selecting a return journey. Prompted with a calendar for this month, I tap on the right arrow to move on to September. About 20 times, my finger hurting more each time. Frustration building significantly, I contemplate giving up, but persevere and finally manage to get the button to accept my prodding.

Selecting September 2nd, I'm presented with yet another list of coaches. Out comes the iPod again so I can check on my return flight time. After scrolling through three or four pages of early coaches I can select one in the mid afternoon that seems like it will suit (again, I wonder what happens if I can't make that coach).

Now that I've selected my ticket I'm offered two pricing options, one an "On-the-day return" ticket and the other the same but with and extra £1 added for insurance. "On-the-day return"? How's that for confusing terminology (it makes me think same-day return), when you've just told the system that you want to return on September 2nd.

Choosing the one without insurance (I'm not going to pay extra out of choice when I'm frustrated by the transaction already) I get an itinerary with a button to press to say I agree to the conditions of carriage. Tapping it, I get another button to click in the corner of the screen to confirm that I'm really happy with the conditions. Not that I had read them, but of course I agreed... I want my ticket, let me pay you already.

The second button tapped and instantly a warning notice appears telling me that I didn't enter my payment information quick enough, even though it didn't actually give me a chance to enter my payment information before telling me I was too slow. Clicking the "yes" button to try again, it gives me a chance to put my card in.

"Authorisation failed." Oh how I wanted to scream. Why do these machines not have alternative payment options so people can pay with cash if desired? Not knowing if there was an issue with my card, I went to try and get some money from the cash machine (successfully) and wandered over to queue up for the ticket desk - something I should have done from the start.

After a few minutes of queueing I was at the desk and talking with a friendly ticketing agent. She took me through where I wanted to go (the terminal question came up again, but she just chose a terminal for me), put me onto a specific coach (but told me it didn't matter which one I took), asked me if I wanted an open return or a specific service (after I asked, she told me the prices didn't differ for that extra level of flexibility) and gave me the ticket. I paid with the same card that the machine had turned down.

As she was taking me through buying a ticket I remarked that I wished the 'express' machines were this easy. She agreed, much like the London Underground ticketing guy did about the problems with Oyster cards on another visit to Heathrow earlier in the year.

I think these machines could be vastly improved upon by simply defaulting to an outbound non-specific coach for the same day and a non-specific coach for the return journey. If people want to change the dates or choose a specific coach then let them, but I suspect the majority of people using those machines are looking to travel now and don't know exactly which coach they'll want to return on. People should be warned that by choosing an open ticket they're not guaranteed a seat, but how often have you tried to travel on a service only to find out that it's full already? Maybe I've only ever travelled on frequent services but I've never come across this problem.

Surely all the system really needs to know is the destination, that you want to travel now and return at some point in the near future and whether you qualify for any discounts (OAP, child, young person's card, etc.). Keep it simple! I get really frustrated when I see technology put in place supposedly to make things easier or faster for people and all they do is make it more stressful, confusing, slower and generally much less satisfactory.

The thousandth link

I posted my thousandth link today to the social bookmarking service del.icio.us.

Red moonThe link itself was Sky & Telescope Interactive Sky Chart which I spotted in Chris Heathcote's links. It caught my attention after the amazing lunar eclipse last night (my blurry photos of the eclipse really didn't do it justice).

To give an idea of the sort of links I bookmark over at del.icio.us, of the 1,000 links over half are tagged with 'geo' (something related to geography), just over 15% tagged with 'drupal' and 7% tagged with 'opensource'. The top 10 are mostly geo-related:

510 geo

164 maps

154 drupal

108 mapping

88 googlemaps

80 geodata

80 germany

79 gis

71 opensource

65 reference

You can view all my public links at del.icio.us/dankarran to see what I've been reading recently (or have saved to remind me to read it later at my leisure).

Gmail logging chat... remotely

I recently gained access to some new features in Gmail that I didn't have available to me before - because I was British.

Now that I claim to American, I can access new features in Gmail such as integrated chat (Google Talk), and even a Delete button. If you're currently set to use Gmail in British English you're probably also missing out on these new features. Set your language to American English and magically they'll appear.

As nicely implemented as the integrated chat is, I found that I would often leave a Gmail window open at work whilst I was out of the office. By doing that, it would accept incoming messages for me before they could reach me at home or on my laptop, meaning that I may not see messages for a day (or actually, in this case, a whole week whilst I was in Canada). Because of that, and my tendancy to carelessly leave various copies of Gmail open, I decided to switch the feature off.

One of the things about the integrated chat is that it logs the messages you send and receive through it, allowing you to search it at a later date. If I was using the integrated client, that would be a very handy feature, an extension of my memory, as Gmail already is (and as del.icio.us is).

What I didn't realise - until today - is that messages are still logged in my Gmail even if I use a separate Jabber client (like Gaim, for example). In my Gmail. That's sort of worrying that this remote service is sitting there, unbeknownst to me, recording my conversations. Admittedly, it is very useful, but it certainly shouldn't do it by default.

Introducing umlauts

When reading an article about Google Earth in Technology Review last month, I came across a new word: co-ordinates.

Obviously being a geographer, that word is not actually a new one to me, but rather a presentation of it that I don't recall ever seeing before. The word is often written as co-ordinates or coordinates, but this time it was written with an umlaut (if that's the English word for two dots hovering over a letter): coördinates. Umlauts are something I see regularly with living in Germany, but not something I would expect to see decorating English words on which I've not previously seen an umlaut sitting.

At the time of reading the article, it puzzled me for a little while, but a transatlantic plane ride quickly helped the puzzling disappear from my mind. On my way back from CeBIT this evening, a friend passed me The New Yorker to help keep me entertained on the journey. To my surprise, once again, I came across this apparent trend in a short column about drinking Tab cola. This time it was with re-engineering (or reengineering), becoming reëngineering.

Has the adding of umlauts to English words been introduced as the new American English way to emphasise the pronunciation of words in which two letters sit together but don't work in the standard way? Is it becoming the replacement for the hyphen in British English, which in the past seems to have largely been omitted from words used commonly in the American lexicon?

Tagging

Just a random thought in a break from revision, provoked probably by spending too much time enjoying surfing the Flickr website for brilliant (and fun) photos...

I've wondered for a while why I've not come across any articles that compare the relatively new phenomenon of tagging things on the internet with the similar - and now largely defunct - concept of using keyword meta tags to categorise web pages. Well, I hadn't read anything until I started writing this, at which point I tryed searching again because I was certain somebody must have compared them at some point. I soon realised that Danny Sullivan of SearchEngineWatch had mentioned it before along with others. There must be some in this long list of related readings that I simply don't have time to read at the moment.

Anyway, I do find it odd that I hadn't come across it before, with most of the things out there being a very positive look at the world of tagging. So far tagging seems to have worked pretty well, though that's how keyword meta tags started out, and look where they are now. I think probably one of the main reasons that tagging is working where meta tags have failed is that tags are in the view of everyone. Meta tags aren't in the public eye, they are hidden away, which meant that web page authors could stuff them with all sorts of irrelevant keywords just to drive up their traffic. If you were to spam the tagspace with irrelevant words, visitors would write the page from the start and find somewhere else.

While Danny Sullivan doesn't see the idea of tags helping searching on the internet, I believe it can and will. For a start, they are the main way of navigating sites like Flickr and del.icio.us, and without tagging these sites would not have grown to the size they are today. They have brought in a new way of navigating the vast amount of information which can soon be generated by a large userbase - but instead of searching, they concentrate on browsing.

Applying tags to the rest of the internet won't work in the same way because it hasn't grown up with these concepts, but I believe they can learn from the information gathered by sites that have employed tags from the start. By learning what tags are relevant to each other (see for example, the related tags list for 'blue' on Flickr: clouds, water, yellow, white, tree, orange, flower, pink, flowers, trees, reflection, window), a search company can start to group sites which are talking about certain subjects. The web page author wouldn't have to do any work at all because the search engines already know what the topic of the page is - this is what search engines specialise in already.

Hmm, conveying that random thought took more of my time than I'd thought, but I should get back to doing some work now. With only one exam down there are two left, but I'm starting to feel a little happier about them now that I've had some of the coursework back from the Easter break and appear to have done quite well.

An interesting use of Amazon Inside data

Amazon concordanceAmazon has quietly introduced an innovative use of their book data - the complete text of many of the books they sell through their website - showing the frequency of the top 100 words inside the book.

Their Concordance functionality is using a technique which has been demonstrated previously on popular tagging-based (or 'folksonomy') sites such as Flickr, del.icio.us and Technorati that shows the relative frequency of each word by varying the font size used to display it. Until now this had only been seen on sites which allow people to categorize things using a number of tags of their choosing, this being the first I am aware of that takes whole texts and picks out keywords in this way.

(via Waxy)

Subscribe to RSS - Information