However, they're delving into what I call "subjective coding" where the choices you're making are no problem at all for a human to decide but they're really inelegant to try and code for.
Reminds me of a recent request I had by a friend. He wanted me to code an application to look through some text documents in which he kept journal entries and find places where he'd made the mistake of substituting a homophone for the correct word, such as "There home is insured" instead of the correct "Their home is insured." And there was one example sentence in which he'd written "I remember one particular incident from my adolescents...", which of course should have been "I remember one particular incident from my adolescence."
I've tried to explain to my friend that to write code to heuristically choose between any one set of homophones would be quite non-trivial. To choose the right homophone among dozens, hundreds?
This is why robots will never take over the world (sorry T3 fans). Programming constraints and processing-speed constraints just won't allow for anywhere near human-speed/human-accurate decisions and judgment.
-kyrathaba
I disagree. I think it's more of a SF/Scifi "racial fear" of letting the robots get too smart, as portrayed in Lowest Common Denominator fashion in the movies.
While I'd certainly grant that it's more than a coding snack, and maybe even for 1 person, it's not "impossible". Quick observation from a slightly spelling-obsessed guy : there are less than thirty "infamous" homophones. There/They're/Their/ ; to/too/two ; lose/loose ; its/it's ; apostrophes vs plurals -- are the top five sets or so.
Then yes, your "adolescence/adolescents" and "intents and purposes (not intensive purposes)" are some more rare ones. But in fact it's "not that hard" because the "spec" called for homophones and suddenly the problem narrows down like a game of 20 questions. Instead of searching every word against every other word, you have a list of all words that CAN be homophones, then when one appears, you check the grammar structure, because it turns out there aren't that many valid grammar structures when homophone pairs are involved.
So yes, it is not a 2 hour lunch snapper, but neither is it in "flying car" territory either. I just believe that we racially have decided that we don't want robots taking over our lives because we haven't socially matured up to Asimov Story level.
Cases in point: The IBM Jeopardy engine is already a quarter up to natural language checking, and what is "cutting edge now" will be "retail" in seven years or so it all goes.
Social commentary time: We had more "fun" blinking about with the Terror/Copyright/BlueBlood theme and wasted a crucial decade and a trillion dollars. Now if that had gone to good ol' social infrastructure we'd be seeing the first wave of fun SF stuff. Instead we're getting the first wave of the dystopian novel visions.