Distributional Inclusion Hypothesis and Quantifications: Probing for Hypernymy in Functional Distributional Semantics
This work addresses hypernymy detection in computational linguistics, offering incremental improvements for FDS models.
The paper tackled the problem of learning hypernymy in Functional Distributional Semantics (FDS) models, revealing that they only learn hypernymy on corpora strictly following the Distributional Inclusion Hypothesis (DIH) and introducing a new training objective that enables learning under reversed DIH and improves detection from real corpora.
Functional Distributional Semantics (FDS) models the meaning of words by truth-conditional functions. This provides a natural representation for hypernymy but no guarantee that it can be learnt when FDS models are trained on a corpus. In this paper, we probe into FDS models and study the representations learnt, drawing connections between quantifications, the Distributional Inclusion Hypothesis (DIH), and the variational-autoencoding objective of FDS model training. Using synthetic data sets, we reveal that FDS models learn hypernymy on a restricted class of corpus that strictly follows the DIH. We further introduce a training objective that both enables hypernymy learning under the reverse of the DIH and improves hypernymy detection from real corpora.