Skip to Main Content
Table 5: 

Performance decomposition to question types of our best generative system (Gen, the best BART-based system), extractive system (Ext, the best BERT-based system, i.e., our best ranker + BERT reader), and ranker (BERT+ICT from Table 2).

Question TypeRatio(%)QA ROUGE-LRanker
GenExtROUGE-L
Relation b/w Concepts 11.0 40.48 24.46 63.76 
Attribute of Concept 12.0 34.09 21.69 56.73 
Event - Attribute 3.4 25.88 10.57 49.23 
Event - Concept 28.3 27.35 15.73 62.15 
Event - Trigger 1.8 29.63 9.28 37.56 
Causal Relation 12.6 22.86 10.39 38.47 
Temporal Relation 12.6 28.01 15.57 49.20 
Nested Relation 15.4 23.02 8.44 48.93 
Book Attribute 2.9 23.11 25.71 54.60 
Question TypeRatio(%)QA ROUGE-LRanker
GenExtROUGE-L
Relation b/w Concepts 11.0 40.48 24.46 63.76 
Attribute of Concept 12.0 34.09 21.69 56.73 
Event - Attribute 3.4 25.88 10.57 49.23 
Event - Concept 28.3 27.35 15.73 62.15 
Event - Trigger 1.8 29.63 9.28 37.56 
Causal Relation 12.6 22.86 10.39 38.47 
Temporal Relation 12.6 28.01 15.57 49.20 
Nested Relation 15.4 23.02 8.44 48.93 
Book Attribute 2.9 23.11 25.71 54.60 
Close Modal

or Create an Account

Close Modal
Close Modal