I've found Shapecatcher (shapecatcher.com) useful here; you draw an arbitrary symbol (e.g. ∑), and it shows you the closest matches from a database of ca. 11000 Unicode glyphs, with each match's code point name (e.g. N-ary summation) and block name (e.g. "Mathematical Operators") -- and once you've found the name of the symbol, you can Google for its meaning.
Thanks, but finding unicode glyphs isn't my problem. I can do that easily enough with the OSX character palette or Julia (which automatically turns TeX escape sequences into unicode characters on the command line). My problem is finding the meaning of a symbol (or composition of symbols -- one above the other, in a subscript, underlined, in a left subscript, etc) in a specific context. Usually what happens when I put a unicode symbol + context into google (or google scholar) I get the worst of both worlds: false positives from the symbol mixed in with the hits I would have gotten with the context terms alone.
I've found Shapecatcher (shapecatcher.com) useful here; you draw an arbitrary symbol (e.g. ∑), and it shows you the closest matches from a database of ca. 11000 Unicode glyphs, with each match's code point name (e.g. N-ary summation) and block name (e.g. "Mathematical Operators") -- and once you've found the name of the symbol, you can Google for its meaning.