3

Reading the news report for the AP Twitter hoax:

Breaking: Two Explosions in the White House and Barack Obama is injured.

I'm thinking this is the most unambiguous and perfect stock market bomb. Is there a measure of how simple this sentence is to parse?

Improve this question
5
  • 1
    What makes an English sentence easier to parse: a single clause; few constituents; constituents consisting of only a single word; and standard SVO word order. There will be other factors.
    – Cerberus
    Apr 25, 2013 at 14:27
  • 1
    As Abney 1995 points out, practically any written English sentence is multiply ambiguous, but native speakers automatically discard unlikely parses for contextual reasons.
    – jlawler
    Apr 25, 2013 at 17:34
  • In the literary sense, besides an implicit phatic or ironic component in a message, veridiction would be outside of this question. A better headline would not include "is". Its unnecessary for journalistic headline but begs the question of its use as the most unambiguous sentence. And as far as this subject any combination of "explosion", "White House", (President) "Barack Obama" could be hard coded bad news (>_<)
    – xtian
    Apr 25, 2013 at 21:43
  • 1
    There will never be such a measure. As jlawler mentioned, there are many levels of ambiguity in NL. There is no absolutely correct way of parsing any sentence. Each grammar formalism and each language model tends to have its own set of preferences for preferring one interpretation over others. Sentences that parse "easily" (i.e. correctly) with one parser, may not be so with another parser.
    – prash ♦
    Apr 26, 2013 at 13:29
  • It might be possible to work out something with Framenet, but detecting this nonsentence as a cue for automatic market algorithms is still in the future. I hope.
    – jlawler
    Jul 22, 2013 at 13:52

1 Answer 1

Reset to default
2

Yes, there are measures. Whether they are consistent or useful is subjective.

A probabilistic parser like https://demos.explosion.ai/displacy or https://cloud.google.com/natural-language/ returns one parse. But under the hood, it had a few candidates and just chose the most probable one.

If the even the first candidate had a low probability then we could say that the sentence was "hard to parse".

If the second candidate had almost the same probability as the first one, likewise.

If the first parse is in fact wrong then tautologically it was (too) hard for that parser.

If it is grammatically correct but semantically improbable then we enter into a philosophical question about what the parsing task includes.

There are also many metrics you can extract from the returned tree itself, like depth.

Improve this answer

Your Answer

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

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