Multi-style Generative Reading Comprehension

ACL 2019 Kyosuke NishidaItsumi SaitoKosuke NishidaKazutoshi ShinodaAtsushi OtsukaHisako AsanoJunji Tomita

This study tackles generative reading comprehension (RC), which consists of answering questions based on textual evidence and natural language generation (NLG). We propose a multi-style abstractive summarization model for question answering, called Masque... (read more)

PDF Abstract

Evaluation results from the paper


 SOTA for Question Answering on NarrativeQA (using extra training data)

     Get a GitHub badge
Task Dataset Model Metric name Metric value Global rank Uses extra
training data
Compare
Question Answering MS MARCO Masque Q&A Style Rouge-L 52.20 # 1
Question Answering MS MARCO Masque Q&A Style BLEU-1 43.77 # 3
Question Answering NarrativeQA Masque (NarrativeQA + MS MARCO) BLEU-1 54.11 # 1
Question Answering NarrativeQA Masque (NarrativeQA + MS MARCO) BLEU-4 30.43 # 1
Question Answering NarrativeQA Masque (NarrativeQA + MS MARCO) METEOR 26.13 # 1
Question Answering NarrativeQA Masque (NarrativeQA + MS MARCO) Rouge-L 59.87 # 1
Question Answering NarrativeQA Masque (NarrativeQA only) BLEU-1 48.70 # 2
Question Answering NarrativeQA Masque (NarrativeQA only) BLEU-4 20.98 # 5
Question Answering NarrativeQA Masque (NarrativeQA only) METEOR 21.95 # 2
Question Answering NarrativeQA Masque (NarrativeQA only) Rouge-L 54.74 # 2