Help Teach a Computer to Recognise Obesity WebsitesMonday, January 16, 2012
There are countless websites out there offering all forms of advice, programs, and services for obesity. Some are run by licensed health professionals, some by governments, some by non-profit organizations, some by self-appointed experts, some by commercial chains, some by scam artists – all of them promise help with managing your weight.
The services range from cosmetic spas, personal coaches, and self-help groups, to registered dieticians, clinical psychologists, and surgeons. Some specialize in exercise, others in diets, some sell dietary supplements, others deliver lean cuisine to your doorstep or offer to surgically implant a gastric band around your stomach.
There are literally 1000s of webpages out there – no one has the time to check them all. Imagine a computer that would help categorize these many sites into commercial and non-commercial, into those offering services by licensed health professionals and those by self-appointed ‘experts’ out to hawk their latest fad diet, into those focusing on psychological interventions vs. those focussing on exercise.
For a human visitor, it actually does not take long to see what a website is about – often one glance is enough to determine where the focus is (e.g. diet, exercise, psychology, dietary supplement, surgery, etc.). A human visitor can often also easily tell if this is a government or non-profit site or a commercial weight-loss program.
These tasks, while easy for humans are actually quite difficult for computers. But, thankfully, computers can be taught and they will learn from ‘experience’.
But teaching a computer requires humans to first tell the computer the right answer. Over time the computer can actually learn to do this by itself – this works for chess, poker, Jeopardy, recognizing faces, and countless other complex processes – so why not for recognizing and categorizing ‘weight-loss’ websites.
The Alberta Centre for Ingenuity and Machine Learning is currently attempting to do just that and this is where you can help.
This link takes you to a random set of about 200 websites that would come up in an internet search on weight-loss. It is now your turn to try and label them into the following categories:
1. Primary type of service that is offered (e.g. alternative medicine, diet, exercise, medical, psychological, etc.)
2. Who offers this service (e.g. public, private, non-profit, commercial, etc.)
3. For whom (e.g. men, women, children, adolescents, etc.)
All you have to do is click on the link – spend a few minutes (this may only take seconds) – and chose the labels that you think are most appropriate for each category. If you are unsure about a site – you can skip it and move on. If more than one label applies (e.g. a site offers both diet and exercise), select multiple options.
Some site may not be relevant to weight management at all – some may not be from Canada (the focus of this project).
Before you start, please spend a few minutes looking at the definitions for each label (on the site) – most are obvious but some are not (e.g. difference between ‘private’ and ‘commercial’ or between ‘physiotherapy’ and ‘rehabilitation’). Simply use your ‘common sense’ to chose the labels that best apply – the computer can only become as smart as you are – you are the teacher.
Remember, you are also not ‘rating’ the sites into ‘good’ or ‘bad’ – you are just sorting them into the right box that best describes what the site offers – so it does not matter whether you think the service is helpful or not – or works or not – or whether you like the information on the site or not – just sort it into the category where it belongs based on what you see on the webpage and move on.
Please label as many sites as you can – every reader will have a different set – so even doing just a few (e.g. the first 20) will create enough information for the computer to work with.
As a word of caution – please do not try to ‘trick’ or ‘game’ the system for e.g. by labelling a ‘commercial’ site as ‘non-profit’ or an ‘alternative medicine’ site as “medical’ – remember there are hundreds of people categorizing these sites and the computer will detect such ‘errors’ and eventually automatically correct them.
Your help is much appreciated.