Densely Connected Attention Propagation for Reading Comprehension

NeurIPS 2018 Yi TayLuu Anh TuanSiu Cheung HuiJian Su

We propose DecaProp (Densely Connected Attention Propagation), a new densely connected neural architecture for reading comprehension (RC). There are two distinct characteristics of our model... (read more)

PDF Abstract
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT LEADERBOARD
Question Answering NarrativeQA DecaProp BLEU-1 44.35 # 3
BLEU-4 27.61 # 2
METEOR 21.80 # 3
Rouge-L 44.69 # 5
Question Answering NewsQA DecaProp F1 66.3 # 2
EM 53.1 # 1
Open-Domain Question Answering Quasar DecaProp EM (Quasar-T) 38.6 # 3
F1 (Quasar-T) 46.9 # 3
Open-Domain Question Answering SearchQA DecaProp Unigram Acc 62.2 # 1
N-gram F1 70.8 # 1
EM 56.8 # 2
F1 63.6 # 3