3
$\begingroup$

FindTextualAnswer isn't working for me here:

txt = CloudImport@CloudObject[
   "https://www.wolframcloud.com/objects/49966c6e-6c9e-4a3e-ba29-b58707ef35ba"];
AbsoluteTiming @ FindTextualAnswer[txt, "AI is finally possible because of?"]

enter image description here

If you ask for the top n most-likely answers, sometimes you get something more reasonable:

enter image description here

What I really need is a probability over these options:

enter image description here

Is there some-work around here to make this work - perhaps comparing embedding distances to each of the answers? Or perhaps someone knows of a multiple choice question-answer networks available in MMA to play with or train?

Improve this question
$\endgroup$
1
  • 1
    $\begingroup$ I think comparing embedding distances is a very sensible start. If you want to do something more sophisticated this paper seems like a good place to begin. $\endgroup$
    – Carl Lange
    Commented Jan 22, 2019 at 8:16

2 Answers 2

Reset to default
2
$\begingroup$ 
txt = CloudImport@
   CloudObject[
    "https://www.wolframcloud.com/objects/49966c6e-6c9e-4a3e-ba29-\
b58707ef35ba"];

choices = {"Big data, Turing test, recommendation test", 
   "Big data, smarter data models, powerful cloud computing", 
   "Cloud computing, general problem solver, paradigm shift", 
   "Big data, mainframe systems, the cloud"};

Simplistically selecting the choice with the largest number of string matches.

SortBy[Transpose@{choices,
    Length@
       StringCases[txt, 
        Alternatives @@
         StringSplit[StringReplace[#, ", " -> ","], ","],
        IgnoreCase -> True] & /@ choices}, Last][[-1, 1]]

(* "Cloud computing, general problem solver, paradigm shift" *)
Improve this answer
$\endgroup$
0
$\begingroup$

With

txt = CloudImport@
   CloudObject[
    "https://www.wolframcloud.com/objects/49966c6e-6c9e-4a3e-ba29-\
b58707ef35ba"];
AbsoluteTiming@
 FindTextualAnswer[txt, 
  "Artificial Intelligence is finally possible because of?"]

you will get a very interesting answer.

Improve this answer
$\endgroup$
2
  • $\begingroup$ I am looking for the likelihood of A, B, C, or D $\endgroup$
    – M.R.
    Commented Jan 22, 2019 at 0:52
  • $\begingroup$ Then you need to change the question. In this formulation, the answer is also funny. $\endgroup$
    – Alex Trounev
    Commented Jan 22, 2019 at 0:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.