We introduce Reprompting, an iterative sampling algorithm that searches ...
Neural sequence generation models are known to "hallucinate", by produci...
Cloud systems are becoming increasingly powerful and complex. It is high...
Pedestrian safety is a priority for transportation system managers and
o...
Current approaches to incorporating terminology constraints in machine
t...
Multilingual pre-trained contextual embedding models (Devlin et al., 201...
While non-autoregressive (NAR) models are showing great promise for mach...
We introduce an Edit-Based Transformer with Repositioning (EDITOR), whic...
While Iterative Back-Translation and Dual Learning effectively incorpora...
Training neural networks with many processors can reduce time-to-solutio...
PubMed is an essential resource for the medical domain, but useful conce...
Natural language understanding in the context of goal oriented dialog sy...
In recent years, with the trend of applying deep learning (DL) in high
p...
Despite some empirical success at correcting exposure bias in machine
tr...
We aim to better exploit the limited amounts of parallel text available ...