Metaphorical Search Engine

http://www.yossarianlives.com
"YossarianLives, a new metaphorical search engine that helps the user discover, create and innovate. YossarianLives is funded by the Technology Strategy Board in the UK and it has been recently featured in the New Scientist, the Guardian and the Wired magazine."  
Website of Yossarian Lives partner and metaphor researcher Dr Ekaterina Shutova: http://www.cl.cam.ac.uk/~es407/index.html ac'd 21/4/14

My first impressions using this is that it misses the delight of metaphor in phonological ambiguities - for example, route didn't give any roots. The idea of conceptualising the metaphor seems counter to the flow of metaphor, and the images, whilst beautiful, often inspiring and occasionally right on target (& thus getting a heart icon response from me) have a fairly similar getty-glow and slick style. What about the more textural gritty side of metaphor? (and creativity!). With some searches the 'conceptual distance' (a slider producing new sets of images closer or further from the search-word) seem more connected than those closer, yet over a few trial searches I can't help but get the impression there are sets of 'people' 'graphs' and not much else that get pulled up regardless. A lot of emphasis in the images on facial expressions and relationships between components of the image.

A really interesting idea and I'll be interested to see how the system 'learns' if I spend longer using it - and how the project develops over time - but for now I'd think of it rather as a paintbox with each pain contained beautifully within it's little shallow palette dish, no messy colours on the lid, and perhaps a paint-by-number style image (but with the numbers encoded for digital scanning by your smartphone - no invisible ink or printed numbers. And the painting would be a summer scene or bunch of Van Gogh sunflowers, on a slightly textured pure white board, A4, just strong enough to stand alone, not strong enough to start cutting into, adding onto, adapting to become more sculptural. 
In other words, a long way to go to begin to approach the rich dynamic and multifaceted world of metaphor elicited with Clean. I'd be interested to know how the attribution of images to the search terms is structured, and the potential to Clean that up!