NLP 2021 @ BGU.CS
Quiz 02e02
Sign in to Google to save your progress. Learn more
Email *
Consider using a Language Model to rank candidates corrections in a spell checker.  The observed non-word is "wird".  The candidate corrections produced by an edit-distance module are ["word", "weird", "wind"].  Write three contexts in which you expect a language model to select each of the three candidates:
Consider a spelling corrector which combines two components: (1) an edit-distance module that proposes candidate corrections based on number of edit operations and their likelihood given the location of characters in the keyboard and (2) a language model to rank candidates based on their probability to appear in the context in which the error is observed.  Give an example where the two models would provide "contradictory" rankings - that is, a case where a word that is "very close to the observed error in edit distance" is ranked very low by the language model in the context, and a case where a word that is "far from the observed error" is ranked high by the context.
Submit
Clear form
Never submit passwords through Google Forms.
This content is neither created nor endorsed by Google. Report Abuse - Terms of Service - Privacy Policy
Google Forms