Don't ‘have a clue’? Unsupervised co-learning of downward-entailing operators.

Cristian Danescu-Niculescu-Mizil and Lillian Lee

Proceedings of ACL, 2010. Short paper.

PDF (with appendices)

Teaser Question:           


One can not answer without knowing whether hon is a “downward-entailing operator” or not. 

If it is, like doubt*, then “We doubt it snowed.” implies  “We doubt it snowed with big flakes.”.

If it is not, like know, then  “We know it snowed with big flakes.” implies “We know it snowed.”

For a language processing system, all words look like hon:  there is no easy way to determine whether a word is downward-entailing operator (and there are many such operators in English and other languages, the vast majority of them missing from the existing hand-constructed lists).

                                                Click here to see the question again.

Talk Slides


System Output

Related Paper:                Without a ‘doubt’



Researchers in textual entailment have begun to consider inferences involving downward-entailing operators, an interesting and important class of lexical items that change the way inferences are made. Recent work proposed a method for learning English downward-entailing operators that requires access to a high-quality collection of negative polarity items (NPIs).  However, English is one of the very few languages for which such a list exists.  We propose the first approach that can be applied to the many languages for which there is no pre-existing hight-precision database of NPIs.  As a case study, we apply our method to Romanian and show that our method yields good results.  Also, we perform a cross-linguistic analysis that suggests interesting connections to some findings in linguistic typology.




  author={Cristian Danescu-Niculescu-Mizil and Lillian Lee},

  title={Don't `have a clue’? {Unsupervised} co-learning of downward-entailing


  booktitle={Proceedings of ACL},



  annote={Short paper}