thunlp/MultiRD

Where does original English Description dataset come from?

Opened this issue · 4 comments

In English experiments, we use the Description dataset from (Hill et al. 2016).

Hill et al. 2016 seems not give an link to download their data, so where does original data come from?

You can email him and he'll give you the link.

Thanks!

Is your data same with original data except spliting into train, test, dev set?

You can email him and he'll give you the link.

Description dataset is exactly the same as his.
The other data is almost the same, except that I've added 22 words' definitions which are in his test-set (200 descriptions) but out of his dictionary definitions by wordnik.

Is your data same with original data except spliting into train, test, dev set?