Monday, 27 September 2010

Non-text search is hard...

As far as I can see, there are basically two types of searchable object:

1. Text objects - such as a web page
2. Non-text objects - such as an image

These two require two different search functions:

1. Text-based search - searches the object itself
2. Metadata search - searches descriptive metadata

In my opinion, all content *should* have descriptive metadata (date written, author, so on). But only non-text objects actually require it - you can't search for 'picture of a horse' unless someone has actively tagged a picture with the word 'horse'.

The problem lies in the subjective nature of descriptive metadata, because tagging/taxonomy-ing/controlled vocabulary-ing is HARD. Really hard. And collecting that data is even harder, and utilising that data is harder again.

Case in point: I have a bunch of hobbies. One of them is perfumes.

If ever there was a set of content that didn't meet with text-based search criteria, it's scent. English doesn't even have scent words, or flavour words - we rely on 'smells like' and 'tastes like' because, outside of sweet/salt/bitter/sour/umami, there are no words (and 'umami', a taste-enhancement molecule found in MSG and parmesan cheese, is stolen from Japanese anyhow).

Because of this, perfumes are the perfect place to examine the efficacy of non-text search. How can I, as a perfume enthusiast and buyer, find new scents?

With difficulty.

Here are some comparisons of my favourite perfumes (I stress: despite what you're about to read, I love these scents), compared against their marketing blurbs.


Hermes Eau de Merveilles

They say:

An enchanting fragrance that encapsulates the spirit of wood, the memory of oceans and the sparkle of a constellation.

I say:

Smells like cake, only sexy. Exactly like I think spice would smell like. (Gentlemen: this is the one you want to buy your lady. You will love this so much.)


Guerlain Samsara

They say:


It was in China and Tibet that Jean-Paul Guerlain sought the inspiration necessary to create Samsara.
Samsara is a Sanskrit word that has its roots in Hinduism. It signifies the cycle of birth and rebirth, the perpetual voyage from one existence to the next, the wheel of life that keeps on turning.

Samsara is an invitation to serenity and harmony, with oneself and with the world. It is a voyage between East and West, Orient and Occident, an homage to the woman who is ever in search of serenity.


I say:

Smells like sand. No organic content, apart from one element - it's sexy (more civet than musk). Exactly like what I think Arrakis would smell like.


Shiseido Zen

They say:

Something to do with amber and 'wood breath'.

I say:

I only bought it because it has space rose in it. Not sexy.


Hermes Kelly Caleche

They say:

The unexpected caress of leather among flowers. Mimosa, iris, tuberose, and climbing rose. A floral perfume touched by leather.

I say:

A wood fire that just went out, and someone dropped some fresh/wet herbs on it. And some booze. Probably gin. Not sexy.


Jardin sur le Nil

They say:

A perfume of light and life, generous and sparkling. A bottle reflecting the shimmering green and ochre tints of the banks of the Nile River appreciated by women and men alike.

I say:

Smells like orange peel and mud, on a very hot day. Not sexy.


Chanel No 5

They say:

Femininity is timeless

I say:

Smells like perfume. Ur-perfume. Super-sexy.


The thing is, I would never search 'orange peel' when looking for a new scent, but neither would I search 'femininity is timeless' or 'cycle of birth and rebirth'. And I know that the results from using a frangrance wheel (yes, that's a Thing) are mid-lame to very-lame.

If I looked for a 'sexy' perfume I would get something vile and musk-ridden (don't fret, it's all synthetic now) with a celebrity branding attached, not something that smelled like cake. Similarly, a search for 'not-sexy perfumes' wouldn't take me to the best not-sexy perfume sources - Japanese houses like Shiseido and Kenzuo and Issey Miyake. (The cultural reasons for this are extensive and, for this post, irrelevant. But Japanese perfumes are amazing.)

Scents - they just don't translate to text.

And yet, a little part of me thinks - hopes - that this is a solvable problem. There will be a way to search for scent. One day.

...and to be honest, pretty much this whole post is to announce how thrilled I am to see that Hermes' new Eau Claire des Merveilles is now out.

Mmmmmmm... fresh cake. Delicious.

Thursday, 9 September 2010

I am so glad this exists: Monocle

I found out about this over a week ago (thank you for making it *and* telling me about it, @josephpearson), so it's about time I posted a synopsis about this potentially excellent new ebook tool.

I haven't played with it yet, but I like the sound of it: it's the Monocle ebook reader. http://inventivelabs.com.au/weblog/post/announcing-monocle/

Monocle is a new - and in my opinion, overdue - development in ebookery. It could be revolutionary, for one reason: it works in browsers. Which is to say, you access your eBook content through the internet.

Fundamentally, there are two ways to access content using machines:

1. Content lives on the user's own device. This is a download-based model.
Example: iTunes.

2. Content lives on external servers which are accessed by the user's device. This is a web-based model.
Example: Spotfiy.

There's a user-facing difference between the two, and a difference in the deployment of IP. In both cases I think the no-download (like Monocle) model will eventually have the upper hand.

For content publishers, the download model is terrifying for a simple reason - once the file is downloaded, it's fate is entirely out of your hands. No matter how much DRM it's wrapped in, someone, someday, will crack it. It will always have the opportunity to go wild, and when it does, there go (some of) the profits.

Similarly, once a file is downloaded, you can't track the usage well. Is it popular? Do people actually use it after they've bought it? It's hard to say (unless the content is software dependant, like iTunes, and that's not much fun).

For content users, the download model is the more annoying option, because it's tethered to a device. We hear people grumble all the time about iTunes, but never about spoftify - spotify is a smoother experience, because there's no fiddling (or protecting) downloaded files to get in the way of accessing the music.

To (shamelessly) use my husband as an example: he works from home some days, from the office others, but he does the same job in both locations. He's an academic, so he uses professionally published academic book content a *lot*.

At work and at home, he uses different machines. They're powerful desktops, so physically lugging them about is not a good option. Because of this, Aadevice-tethered eBook is no use to him.

But he always has an internet connection. He can always log on to a website.

There's a developing trend of people using multiple machines to access their content, instead of the trusty old family PC. My household isn't unusual - it has two people, a desktop, two laptops, an ipad, and an iphone. Matching that trend on the other side, we can also see internet access growing exponentially - it's quite likely (if not inevitable) that ubiquity is just around the corner.

Which is more likely? A future where people use the same device all the time, or a future where people have internet access all the time?

In choosing between browser-based and download-based content models, these respective trends point to an access-it-through-the-internet model as being where the smart money is.

A browser-based service naturally has to rely on a logon. This for me is a double-bonus, because (unless I've misinterpreted) Monocle can support the standard university logon services such as Raven and Shibboleth. That's great for universities, because that opens up the ability for a university library to buy browser-based eBooks and make them accesssible to all staff and faculty using their own, normal logon - no mess, no fuss, all access.

Monocle is also (hooray!) completely open source. I hope it does really, really well.