Kirstain et al. proposed s2e (start-to-end) coref model, a simple and efficient transformer-based coref model,
to mitigate large memory footprint
and achieved three times better memory efficiency.
Authors
Yuval Kirstain, Ori Ram, and Omer Levy (Tel Aviv University)
Motivation
The previous method uses span representation to perform coref, which requires O(n2d) space and much GPU memory.
Method
The proposed method uses only span boundaries (ie, its start and end tokens); here there is no representation for each span (but just for boundaries), leading smaller space.
Results / Insight
Three times more (GPU) memory efficient, without deteriorating the performance.
Coreference Resolution without Span Representations (ACL21)
Contribution summary
Authors
Yuval Kirstain, Ori Ram, and Omer Levy (Tel Aviv University)
Motivation
Method
Results / Insight