RST Parsing on RST-DT original Parseval (test)

78.7Span F1

Human

Updated 1mo ago

Evaluation Results

Method
Human 2021.02	78.7	66.8	57.1	55
DeBERTa 2022.10	78.5	67.9	56.6	54.4
XLNet 2022.10	77.8	67.4	57	54.8
DeBERTa 2022.10	77.8	68	57.3	55.4
RoBERTa 2022.10	77.3	66.6	55.8	53.8
SpanBERT 2022.10	76.5	65.4	54.5	52.2
Zhang et al. 2022.10	76.3	65.5	55.6	53.8
RoBERTa 2022.10	76.1	66.5	55.4	53.7
XLNet 2022.10	76.1	65.9	56.3	54.2
SpanBERT 2022.10	76.1	65.3	54.9	52.7
LSTM (dynamic) 2021.02	73.1	62.3	51.5	50.3
LSTM (static) 2021.02	72.7	61.7	50.5	49.4
Yu et al. (2018) 2021.02	71.4	60.3	49.2	48.1
Yu et al. (2018) 2021.02	71.1	59.7	48.4	47.4
Transformer (static) 2021.02	70.6	59.9	50.6	49
Transformer (dynamic) 2021.02	70.2	60.1	50.6	49.2
BERT 2022.10	69.8	59.1	48.3	46.6
Feng and Hirst (2014) 2021.02	68.6	55.9	45.8	44.6
BERT 2022.10	68.3	57.8	47.8	46
LSTM (dynamic) (no paragraph feature) 2021.02	67.3	57.4	48.5	47.4
Zhang et al. (2020) 2021.02	67.2	55.5	45.3	44.3
LSTM (static) (no paragraph feature) 2021.02	66.3	56.6	47.1	46.1
Surdeanu et al. (2015) 2021.02	65.3	54.2	45.1	44.2
Joty et al. (2015) 2021.02	65.1	55.5	45.1	44.3
Hayashi et al. (2016) 2021.02	65.1	54.6	44.7	44.1
Li et al. (2016) 2021.02	64.5	54	38.1	36.6
Ji and Eisenstein (2014) 2021.02	64.1	54.2	46.8	46.3
Braud et al. (2017) 2021.02	62.7	54.5	45.5	45.1