Conversational Discourse Parsing on STAC (test)
74.1Arc ScoreDiscProReco(w/o DPR)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DiscProReco(w/o DPR)with DPR=false2021.06 | 74.1 | 57 | — | — | |
| Deep Sequential2021.06 | 73.2 | 55.7 | — | — | |
| Deep+MSTArchitecture=GRU + MST2021.06 | 69.6 | 52.1 | — | — | |
| Deep+GreedyDecoding=Greedy2021.06 | 69.3 | 51.9 | — | — | |
| Deep+ILPArchitecture=GRU + ILP2021.06 | 69 | 53.1 | — | — | |
| MST2021.06 | 68.8 | 50.4 | — | — | |
| ILP2021.06 | 68.6 | 52.1 | — | — | |
| Deep Sequential ParserTraining data=Joint (STAC+Molweni)2021.10 | — | — | 72.8 | 54.8 | |
| Our Proposed Parser w/ language backboneTraining data=Joint (STAC+Molweni), Backbone=RoBERTa-base2021.10 | — | — | 75.5 | 57.2 |