3
$\begingroup$

I have a text string containing unstructured data and I would like to analyze it in order to extract structured information. In particular, this text string specifies when a service is operational (the days and the hours). This text string may be written in different ways:

  • a list containing one or more day names;
  • hours may be written next to the names of the days to which they refer;
  • the string may contain one or more range of days (each range of days could have a certain range of hours);
  • the name of the days may not be present, and in this case it could be replaced by "every day" or the words "workdays" or by "holidays";
  • etc..

What approach could I use to be able to extract structured information from such a text string? Are there specific algorithms to achieve this purpose? Should I use a classifier (artificial neural networks, random forests, etc.)?

$\endgroup$
1
$\begingroup$

What approach could I use to be able to extract structured information from such a text string?

The task corresponds to named-entity recognition + relation extraction. Two main choices:

  • Unsupervised, which in your case most likely mean defining regular expressions and a bunch of heuristic ( e.g., hours directly succeeds days to reach that refer to )
  • Supervised

You can mix both choices, e.g. perform the named-entity recognition supervisedly, and relation extraction unsupervisedly.

From the examples you gave, I'd go unsupervisedly at first.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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