HyperAIHyperAI超神経

Command Palette

Search for a command to run...

プラットフォーム

Discourse Parsing On Rst Dt

評価指標

RST-Parseval (Nuclearity)
RST-Parseval (Relation)
RST-Parseval (Span)

評価結果

このベンチマークにおける各モデルのパフォーマンス結果

Paper Title
Bottom-up Linear-chain CRF-based Parser71.058.285.7-
Top-down (XLNet)---A Simple and Strong Baseline for End-to-End Neural RST-style Discourse Parsing
Top-down Llama 2 (13B)---Can we obtain significant success in RST discourse parsing by using Large Language Models?
Top-down (SpanBERT)---A Simple and Strong Baseline for End-to-End Neural RST-style Discourse Parsing
Guz et al. (2020)---Unleashing the Power of Neural Discourse Parsers -- A Context and Structure Aware Approach Using Large Scale Pretraining
Re-implemented HILDA RST parser66.6*54.6*82.6*-
Greedy Bottom-up Parser with Syntactic Features67.1*55.4*82.6*-
HILDA Parser68.455.383.0A Novel Discourse Parser Based on Support Vector Machine Classification
Bottom-up Llama 2 (70B)---Can we obtain significant success in RST discourse parsing by using Large Language Models?
LSTM Dynamic---Top-down Discourse Parsing via Sequence Labelling
Two-stage Parser72.459.786.0A Two-Stage Parsing Method for Text-Level Discourse Analysis
Bottom-up (DeBERTa)---A Simple and Strong Baseline for End-to-End Neural RST-style Discourse Parsing
DMRST---Bilingual Rhetorical Structure Parsing with Large Parallel Annotations
LSTM Sequential Discourse Parser (Braud et al., 2016)63.6*47.7*79.7*Multi-view and multi-task training of RST discourse parsers
Transformer (dynamic)---Top-down Discourse Parsing via Sequence Labelling
Bottom-up Llama 2 (13B)---Can we obtain significant success in RST discourse parsing by using Large Language Models?
Top-down Span-based Parser with Silver Agreement Subtrees74.762.586.8Improving Neural RST Parsing Model with Silver Agreement Subtrees
Bottom-up (RoBERTa)---A Simple and Strong Baseline for End-to-End Neural RST-style Discourse Parsing
End-to-end Top-down (XLNet)76.061.887.6RST Parsing from Scratch
Bottom-up (SpanBERT)---A Simple and Strong Baseline for End-to-End Neural RST-style Discourse Parsing
0 of 40 row(s) selected.