Slashdot Mirror
Fighting Food Poisoning In Las Vegas With Machine Learning
aarondubrow writes: Computer science researchers from the University of Rochester developed an app for health departments that uses natural language processing and artificial intelligence to identify likely food poisoning hot spots. Las Vegas health officials recently used the app, called nEmesis, to improve the city's inspection protocols and found it was 63% more effective at identifying problematic venues than the current state of the art. The researchers estimate that if every inspection in Las Vegas became adaptive, it could prevent over 9,000 cases of foodborne illness and 557 hospitalizations annually. The team presented the results at the 30th Association for the Advancement of Artificial Intelligence conference in February.
Do we really need an artificially intelligent app to tell us not to eat the comp'd seafood?
It ruins the experience if I can't let the world know.
It depends how consistently and well you describe adverse occurrences. Yawn.
Those who profit from food poisoning (health department) are not interested in making detection better.
This should be marketed to consumers, who don't want to get sick.
Or course, it is just a matter of time before restaurants start paying protection money to never be listed in the app.
How dare they use complaints to find and shut down the business of hard working food poisoners!
excitingthingstodo.blogspot.com
The cruise ships should get some benefits from this though.
“He’s not deformed, he’s just drunk!”
When traveling to Las Vegas, avoid Chipotle's.
What if this ends up unintentionally profiling some protected group? Maybe some culture's traditional food preparation methods are statistically less hygienic or what have you, or make more use of raw and under-cooked meet or eggs or similar things?
We must all unite and fight this.
sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
I remember the University of Rochester was known for having one of the leading AI and optics research departments in the US . Kudos to them here.
"Imagination is more important than knowledge" - Einstein
emesis /emsis/
noun technical
the action or process of vomiting.
Someone has a sense of humor at the University of Rochester!
Horror & SciFi Erotic Nudes
Basing your food poisoning adaptive learning system off of tweets? You might as well base your understanding of the cosmos on the drunken hobo that says he was abducted by aliens and taken to their home world. Sure its probably better than random chance but so are tea leafs. You might be better off making your own app (or cooperating with social media app makers) to create some app that somehow anonymizes say a weeks worth of locational information and lets a person note their health status (Little headache, sick in bed, I think I'm dying, etc). When a person starts reporting symptoms it tracks back where they've been for the past week and see's if it coincides with other people who are reporting symptoms.
It's las vegas. Everyone there believes they can beat the odds. Come on, try the suchi. What are the odds you'll get sick? Are you chicken, Colonel Sanders?
"Transparent" is a shit show that trades on every stereotype going. A man in drag is NOT a transsexual.
>The most important distinction, however, is between
restaurants with minor violations (grades A and B) and those
posing considerable health risks (grade C and worse). nEme-
sis uncovers 11 venues in the latter category, whereas control
finds only 7, a 64% improvement.
They found 4 (count 'em FOUR) more restaurants than the control and conclude
>Given the ambiguity of online data, it may appear hope-
less to identify problematic restaurants fully automatically.
However, we demonstrate that nEmesis uncovers signifi-
cantly more problematic restaurants than current inspection
processes
FOUR more than the control is NOT significant in light of the fact that they inspected 142 venues and found only FOUR more violations than the controlled experiment AND they cite the fact that they found an illegal venue running without a permit that would've never been inspected anyway. (Which is both a pro or con depending on how you look at it - pro because it wouldn't have been found otherwise, con because that's a problem that should be discovered by any beat cop on the strip.
Lastly, the entire concept relies on social media and voluntary data collection which is incredibly flakey and prone to social attitudes of the day. I bet this system would've worked even better using foursquare (maybe even yelp) but those apps have fallen by the wayside and twitter will too. What's the critical mass that requires statistical suppositions to target a potential venue for an inspection? 5 complaints? 2? 1? At that point the process becomes subject to trolling and hacking. (Let alone generational and geographical effects about the type of tweets that get posted.)
It's a novel approach - but it's long term effectiveness is highly doubtful.
You want to avoid the "Space Special" too.
ELOI, ELOI, LAMA SABACHTHANI!?
Wouldn't this all be much simpler if we just made the hotels and casinos pay the affected guests' hospital bills? And if necessary add punitive fines so that poisoning guests is more expensive that just paying kitchen workers a decent living wage, keeping the kitchens clean, and following basic food safety rules.
"...uses natural language processing and artificial intelligence to identify likely food poisoning hot spots."
I don't need an app for that. It's the free buffets in casinos and strip clubs where people touch the serving utensils with their (unwashed after bathroom use) hands and where droplets of sniffles, coughs and sneezes are spread over the food.
I just can't imagine... something tells me that immigrants from festering, third world hellholes have very poor hygiene standards...
But that would be 'racist' - who cares about people suffering the agony (and sometimes death) caused by food poisoning?