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

TL;DR

This paper introduces an unsupervised co-learning approach to identify downward-entailing operators in languages lacking high-quality NPI lists, demonstrated on Romanian, with implications for linguistic typology.

Contribution

It presents the first unsupervised method for learning downward-entailing operators applicable to multiple languages without pre-existing NPI databases.

Findings

Method yields good results on Romanian

Cross-linguistic analysis reveals typological connections

Approach extends inference capabilities across languages

Abstract

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 high-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.