Okay, this is officially bad science week at Rough Theory… For those who haven’t seen this yet, I wanted to link to an article at TechCrunch on healthBase, which offers “Answers automatically retrieved from millions of authoritative health sources”. HealthBase’s search field invites you to “Enter a health condition, disease or sign”, so that its semantic search algorithms can aggregate information from across its various sources, in order to present to you a contextually appropriate list of treatments, causes, complications, and pros & cons of specific treatments.
Leena Rao from TechCrunch reports that, when users began exploring the new search engine, they found some… unexpected results. Specially, she reports that healthBase lists “Jew” as one of the causes of AIDS (please check the screenshots from the original piece). It gets better:
When you click on Jew, you can see proper “Treatments” for Jews, “Drugs And Medications” for Jews and “Complications” for Jews. Apparently, “alcohol” and “coarse salt” are treatments to get rid of Jews, as is Dr. Pepper! Who knew? I’ve included the screenshots of the results below if you don’t believe me.
The answers don’t seem motivated by anti-semitism, but by content aggregation that doesn’t adequately parse how the meaning of words can change in different contexts. Thus, as Rao reports further down, you get other, less charged, but equally nonsensical, results from other searches:
If you look at the pros of AIDS (yes, it thinks here are pros to having AIDS), it comically lists the “Spanish Civil War.”
Those responsible for healthBase are working on the problem and promise rapid improvements. Meanwhile, friends to whom I’ve passed on the link keep spamming me with their own favourite results. I’ve also enjoyed the results posted at jonquil’s site, where I originally saw the link to the TechCrunch article.
Edited to add: just to give a taste of what gets generated, I’ll include the result I added to the discussion over at Wildly Parentherical:
I got “pain” as a complication of “dying”…
Treating dying has certain pros, including: “achieve commercial success”, “fine surface finish”, “pioneer hip hop music”, and, my personal favorite: “add bonus die” (which, I suppose, is one way you could look at being brought back to life, if you were so inclined…)
If you’d rather not treat dying, perhaps you would prefer to avoid its causes, the first of which is apparently “ginger”…
See further commentary at davidrothman.net.
[...] How content aggregation explains that the upside of contracting AIDS is “The Spanish Civil War… seen via tnh’s [...]
Masculinity – Treatment: Discourse analysis
lol – yeah, I see that one of the complications of “heterosexuality” is “myth”.
Its treatments include “homosexuality”, “Nineteen eighty four”, “Qualitative research design” ;-P, “Age” and “Appearance”…
It has only one listed cause, which is “force”…
Its pros include “Encompass entire social identity”…
Well, Japanese caused cholera, according to their search engine. So there you go – covering all the bases.
Drugs and medications for spam: Viagra
Now JB: you know what happened to your comment when you first tried to post it, right? :-) (Here’s hoping that pulling you out of the spam bucket doesn’t unleash a flood of Viagra spam onto the blog… ;-P)
As the original inventor of the company’s NLP technology (but two years removed from the company), I have broken down the causes of the bizarre (and amusing) errors. Enjoy the funny stuff while it’s there because the problems can and should be fixed in due time:
http://nlpconfidential.blogspot.com/2009/09/medicine-for-healthbase.html