1
$\begingroup$

I am unclear about the practical benefit of named entity recognition. Specifically, I don't get how this entity is any better than a string representation of the same object. Let's take the popular example of movies. I can create a model where each row is a movie. Any attribute you can have in a recommendation system can just as easily be a column in my movie table: release date, actors, genre, etc. If a user does a search for 'animation' - or even if they don't, but you know from that user's history that they saw The Lion King, any decent search engine could pull up Moana as a recommendation, so how is a named entity any better?

Or you pull a named entity out of a text document. Great, but now what? Isn't it still just a string, which can be assigned as an object with it's own unique dict of values, like genre and release date? I assume there are benefits over strings, or it wouldn't be such a big deal to so many people. I just don't know what those benefits are.

$\endgroup$
0
$\begingroup$

The comparison between named entity and string is irrelevant, because the concept of named entity doesn't belong to data types such as string or numeric values. Of course a named entity is usually represented as a string but so is a text document, a cryptographic key or a translation of a text to Vietnamese: what matters here is the semantics, not the technical representation.

The reason why named entities (NEs) are often offered special treatment is because they are particularly relevant for certain tasks and are notoriously difficult to detect and analyze:

  • The task of Named Entity Recognition (NER) consists in detecting named entities in a raw text document. It's not a trivial task, the problem is still an active area of research.
  • Even after having been identified, named entities are often difficult to disambiguate. The same entity may appear under many different variants for various reasons:
    • "John Smith", "J. Smith" and "John A. Smith" may or may not be the same person.
    • "The president" and "Donald Trump" may or may not be the same entity.
    • "Big Apple" and "New York" represent the same entity, but not "big apple" and "New York".
    • ...

Of course when one is provided with the result of the NE recognition/disambiguation process, they can process NEs like any other piece of information... but that's because all the hard work about NEs has been done before.

$\endgroup$
  • $\begingroup$ The links you provided were very helpful, in addition to your answer itself. Thank you! $\endgroup$ – Malik A. Rumi Jul 29 at 15:47

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.