Tea and Pattern Recognition

A post in the ‘Meet The Team’ series.

I’m Sarah Harry.  I’m a clinical terminology specialist working with SNOMED CT to build knowledge representations in the world of clinical laboratory medicine that everyone can share intelligibly.  ‘Terms, Lists, Sorted’ is actually my job.

You will have heard of Deep Blue, a super-computer that beat the leading chess player of the time.  Subsequent iterations have also beaten the best Go player too.  Deep Blue could process 200 million board positions every second.  A human being can barely process one. But a human being can process those perhaps 100,000 positions it does know in parallel.  It’s what makes the brain the most powerful thing in the universe.  We are slow but massively parallel.  Our brains don’t have much more of a ‘configuration file’ than other apes and the structures are just huge numbers of the same canonical brain cell form in the neocortex but it allows all that parallel analysis.  Computers will achieve this power too but not yet.  And what is the basis for how this parallel processing comes to conclusions?

The parallel processing uses pattern recognition.  Pattern recognition is every human being’s superpower.  We all do it every moment of the day.  The philospher Isaiah Berlin may have said: ‘To understand is to perceive patterns. To make intelligible is to reveal the basic pattern.”  Tiago Forte describes it thus with reference to Ray Kurzweil’s ‘Pattern Recognition Theory of Mind:

“What the mind is doing when it ‘recognises’ an image is not matching it against a database of static images. There is no such database in the brain. Instead, it is reconstructing that image on the fly, drawing on many conceptual levels, mixing and matching thousands of patterns at many levels of abstraction to see which ones fit the electric signals coming in through the retina.”

I spend a lot of my time working from existing knowledge hierarchies and unstructured terms and lists and I use my internal conceptual hierarchy to reify these into structured concepts in SNOMED CT that represent requests and results.  In the process I am both analysing from my existing pattern knowledge and updating it.  These patterns too have their hierarchies and re-form upwards and downwards as content formation for the domain progresses.  I separate noise from signal, inhibiting pattern application that I don’t expect to see but this can also cause me to lose important data as we all see what we expect to see and screen out what is judged as ignorable.  If we didn’t, life would be impossible.  It needs care; we are so good at patterns we even see patterns where they don’t exist!

The domain specific knowledge lists I’m working on are in their turn recognisable by laboratory scientists and clinicians, applying their own pattern recognition systems to my regularised terms.  Servers and applications, in their cruder way, apply semantic recognition and content annotation to the queries and responses of the human users as we build in algorithmic AI and machine learning to our interfaces, facilitating code list use by labs, clinicians and patients. 

While my brain is doing this amazing ‘massively parallel pattern recognition’ thing, I’m just staring at lists of laboratory test terms trying to decipher strange local shorthand abbreviations and drinking massive amounts of Yorkshire Tea.  And in the words of designer Phamie MacDonald: “Without tea there is only darkness and chaos.”

Lots on the Internet on ‘humans as pattern seekers’.  Enjoy!