This is a lighthearted romp while I am watching the cricket…
Sulfur chloride:
WA knows that it’s ambiguous – that could be lookup – but the IUPAC name ?????. I want to see Daniel’s expression when he sees it – he has a splendid way of indicating when he feels a name or structure violates the rules. I couldn’t find the “IUPAC” name in Pubchem or elsewhere so maybe Wolfram is generating it. If so, it’s going to have to encode an awful lot of rules.
Pentachloromethane (non-chemists – this doesn’t exist – carbon has only 4 valencies). However WA makes a brave (and completely wrong) guess:
So it looks like it has some sort of natural language engine. I’m not sure that’s a good thing to mix with algorithmic reasoning
It gets phosphorus trichloride and phosphorus pentachloride right, so I tried a non-existent compound.
We should remember that this is an Alpha, so it will have mistakes. Since it’s Closed we have to use these sorts of tricks to try to find out how it thinks. I’ll finish with our old friend:
This is internally inconsistent. The “formula” doesn’t match the Structure Diagram. It’s carefully put brackets into the name and then interpreted it wrongly. Maybe it’s a clever algorithm which just needs tuning or maybe it’s a fuzzy approach which is struggling.
However they have clearly paid people to put data in. I wonder where from?
At the bottom of every page, they have a link called “Source information”. In general, it comes from Mathematica’s ChemicalData module (http://reference.wolfram.com/mathematica/note/ChemicalDataSourceInformation.html) with some natural language processing thrown in, as you’ve pointed out. (In mathematica, the module acts like a perl hash, if there’s no key for the name you type in, you don’t get anything back, so none of this pentachloromethane stuff would work.
@David Thanks – I will have a look