Examples of sentence pairs with bi-modal human judgment distributions. Examples are drawn from SNLI, the VerbNet portion of DNC, and the MegaVerdicality portion of DNC (from left to right). Training distribution is in blue; test in orange. Dotted black line shows the model fit when using a single component; shaded gray shows the model learned when allowed to fit k components. Distributions are over z-normalized scores in which 0 roughly corresponds to neutral (p ↛ h) but not precisely (§3.3).
This site uses cookies. By continuing to use our website, you are agreeing to our privacy policy. No content on this site may be used to train artificial intelligence systems without permission in writing from the MIT Press.