Sense-enhanced embeddings produce graphs that better capture semantic type information, allowing Neighbor Type Probability and Neighbor Type Entropy metrics to distinguish type matching from coercion in contextualized noun embeddings.
Unsupervised Content Discovery from Concise Summaries
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
A graph-based analysis of semantic types and coercion in contextualized word embeddings
Sense-enhanced embeddings produce graphs that better capture semantic type information, allowing Neighbor Type Probability and Neighbor Type Entropy metrics to distinguish type matching from coercion in contextualized noun embeddings.