FindTextualAnswer

FindTextualAnswer[text,"question"]

gives the substring of text that best appears to answer question.

FindTextualAnswer[text,"question",n]

gives a list of up to n answers that appear most probable.

FindTextualAnswer[text,"question",n,prop]

gives the specified property for each answer.

Details and Options

  • In FindTextualAnswer[text,], text can be a string, a file with plain text, a ContentObject expression or a list of these text objects.
  • FindTextualAnswer[text,{"question1","question2",},] gives answers for each of the questioni.
  • The property prop can be one of the following:
  • "String"substring corresponding to candidate answer
    "Probability"
  • estimated probability that the answer is correct
  • "Position"
  • start and end positions in text of the answer
  • "Sentence"sentence containing the candidate answer
    "Line"line containing the candidate answer
    "Paragraph"paragraph containing the candidate answer
    "HighlightedSentence"sentence with candidate answer highlighted
    "HighlightedLine"line with candidate answer highlighted
    "HighlightedParagraph"paragraph with candidate answer highlighted
    {prop1,prop2,}a list of property specifications
  • The following option can be given:
  • TargetDevice"CPU"target device on which to perform the computation
    PerformanceGoalAutomaticwhich aspects of performance to optimize
  • Possible settings for PerformanceGoal include "Speed" and "Quality".
  • FindTextualAnswer uses statistical methods to try to find answers. Details of the text and question can affect its performance.
  • FindTextualAnswer may download resources that will be stored in your local object store at $LocalBase, and that can be listed using LocalObjects[] and removed using ResourceRemove.

Examples

open allclose all

Basic Examples  (2)

Answer a question from a sentence:

Answer a question from a paragraph:

Scope  (6)

Find multiple possible answers from a text:

Find multiple possible answers along with their probability:

Answer several questions:

Find the sentence containing the answer:

Get the position of the answer in the question:

Answer several questions and display the answers in context:

Look for the answer to a question from a ContentObject expression:

Options  (2)

TargetDevice  (1)

Answer questions using the default system GPU, if available:

The computation is typically faster than with CPU:

If a compatible GPU is not available, a message is issued:

PerformanceGoal  (1)

Use PerformanceGoal"Quality" to get a more accurate answer:

Applications  (1)

Search an Answer to a General Question in Wikipedia  (1)

Display in context the possible answers from a Wikipedia search:

The instances of the same names can be grouped along with their probabilities:

Then the one with the largest total sum of probability is probably the good answer:

Possible Issues  (1)

FindTextualAnswer is able to understand semantic concepts to some extent but is not able to use logic to guess a number:

Introduced in 2018
 (11.3)
 |
Updated in 2020
 (12.1)