Friday, 10 April 2009

I made a wonderful discovery yesterday.

I was cycling home after some volunteer work at a local primary school (nb: my day usually doesn't resemble Mary Poppins' quite so closely) when I accidentally found a new aquarium store: http://www.cambridgecoraltech.co.uk/.

I have dreamed of a store like this. Never thought I'd find one.

They sell coral. They sell lovely normal tropical fish. They also sell banded coral shrimp, lionfish, seahorses, pipefish, stonefish (OH MY GOD), stingrays (OH WOW) and - are we ready? SEA JELLIES.

Sea jellies! No way.

Sea jellies are super-hard to keep. They need constantly moving water and very specific foods. (The jellies on sale are apparently quite easy, becuase they like to lie upside down on their bells with their tentacles floating upwards. Nature, eh?)

Now why didn't I find this store before?

Because the browse mechanism that lead me to it was 'cycling down the right street'.

I've been intrigued for a while now by how bad conventional search is for finding local businesses. Location specificity in search is being improved by google maps, but there's a long way to go.

What excites me about improvements to location based search is this: That it will have a huge, and positive, impact of local small business. If a quick search can reveal that there is a specialist alternative to B&Q or homebase nearby, then that is more business that the specialist can potentially get, which in turn will lead to more specialists.

I could be wrong. But I hope that, if location based search gets really good, we could have a small-business renaissance. And wouldn't that be nice?


Ambiently

I quite like this.

In a conventional, keyword based search engine (I'm looking at you, Google) there's a lot of guess-work involved. It works like this:
  • There is a 'thing' out there in the world that can be described many ways using 'words'. (And yes, this remains true when the 'thing' is made of 'words', as the majority of open-web 'things' are.)
  • There is the author of a website. She describes her 'things' using certain 'words'.
  • There is a user. She must guess which 'words' the author used to describe the 'thing'.
For top-level concepts, that works fine. Take the educational content-packaging standard called 'SCORM'. Search using word SCORM, and you'll find a bunch of sites about SCORM. So far so good.

But if your search is more specific, it gets tricky. Let's say the user has a SCORM package and she'd like a free piece of software to test it for compliance to the SCORM 2004 specification. Now she has to guess words: 'free' 'compliance' 'SCORM' '2004' (bearing in mind that google uses whitespace as a delimiter - i.e. a way of breaking up English into meaningful chunks). Lots of results, but no testing tool! So she guesses again. And again. And again.

Ambiently provides a simple, but potentially very poweful, change. Via a button on the toolbar, it will present the user with 'related websites'. Instead the user guessing keywords, ambiently (I'm guessing based on my own limited trials, here) uses the site you're currently in as keywords. Start in the almost-right-place, and it will show you what's close. No guessing.

It's a small change. It's not semantic (my standard test of semantic value - 'piracy' - returned as much on copyright as it did on high seas crime). You won't see substantially different results from google. But I suspect that for intricate and/or highly specific searches, it will prove extremely useful.

Also - it brands itself as a 'discovery' engine, rather than a 'search' engine. Full props to that - it's about time we stopped using 'search' as a catch-all term for discovery. 'Search' implies that you already know exactly what you're looking for, denying the value of serendipitous discovery. Sure, it's semantics, but semantics are the bread and butter of content discovery!

www.ambiently.com

...and the answer to the question is: there is an excellent free SCORM compliance tool at www.scorm.com. If you use SCORM, this site is invaluable.

Sunday, 5 April 2009

Unique identifiers that aren't

...This has been bothering me for some time.

Here are my top two contenders for the most irritating 'unique identifiers that aren't'. By that I mean: terms that should refer to only one entity within a range of entities, but don't. 

If you search for 'Apple' on the open web (e.g. in Google), then yes, you will get a combination of fruit and tech. That's annoying, but the two entity sets - 'fruit' and 'tech' - are fundamentally different. If you move to an environment that serves a specific entity set - a supermarket online service, or cNet - that if you search 'apple' you'll get the relevant entity.

But what if there were two tech companies called 'Apple'? Entities with the same name in the same entity set? Well, of course, there are. Here are my least favourite:

1. Place names. 

I live in a town called 'Cambridge'. It is in the UK. There is also a town called 'Cambridge' in Massachusetts, USA. More annoyingly, they both have prestigious universities (in the US, Harvard). Searches on the open web usually return both.

It's not a big problem for me, but I have undying sympathy for the good people of Paris, Texas, and Melbourne, Florida. (The Paris people especially - they must curse that movie).

Still - country filters usually sort this one out without much fuss. Unlike...

2. People names.

My name is Zoe Rose.  A google search for "Zoe Rose" (mini google hack: double quotes tell the engine to search for the words as a string, i.e. disallowing the whitespace delimiter) will return, in reverse order:
  • Style A to Zoe by Rachel Zoe, Rose Apoda Apodaca by Little...
  • Zoe Rose Mogan Jennings
  • Zoe Rose - London | Facebook
  • Motel @ Topshop Urban Outfitters Zoe Rose Dress M 10 12 on eBay
  • Zoe Rose Designs - Hereford textile artist creating handmade...
  • Video result: Piedemont High School Senior Speaker Zoe Rose
  • bebo.com - Profile from Zoe Rose.
  • bebo.com - Profile from Zoe Rose
aaaaand, my personal favourite, result number one:
  • Zoe Rose - Actress. World Sex Tour 23: England. Visit IMDn for photos, filmography...
That's right. A porn star stole my name. 

You have to get to the third page before 'Zoe Rose' meaning me comes up at all - and the reference is way out of date. While it's nice to have some privacy, there is something cringeworthy about being outranked by a dress. 

So let's disect. My unique identifier is muddled up with: 
  • People who have part of my name - google can't differentiate between 'Rose' as middle name and 'Rose' as surname
  • Inadvertent co-location of 'Zoe' and 'Rose', due to rose being a proper noun as well as a name
  • People who legitimately have the same name as me (fair enough)
  • Entities that aren't people at all (the dress)
  • A porn star who I really, really wish had picked a different pseudonym
I can't think of a solution to this one. But I really hope someone else does - and soon.