Document Summarization On Cnn Daily Mail

評価指標

PPL

ROUGE-1

ROUGE-2

ROUGE-L

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

モデル名	PPL	ROUGE-1	ROUGE-2	ROUGE-L	Paper Title	Repository
C2F + ALTERNATE	23.6	31.1	15.4	28.8	Coarse-to-Fine Attention Models for Document Summarization	-
Lead-3	-	40.34	17.70	36.57	Get To The Point: Summarization with Pointer-Generator Networks	-
Hie-BART	-	44.35	21.37	41.05	Hie-BART: Document Summarization with Hierarchical BART	-
TaLK Convolutions (Deep)	-	40.59	18.97	36.81	Time-aware Large Kernel Convolutions	-
Scrambled code + broken (alter)	-	48.18	19.84	45.35	Universal Evasion Attacks on Summarization Scoring	-
BERTSUM+Transformer	-	43.25	20.24	39.63	Fine-tune BERT for Extractive Summarization	-
ML + RL (Paulus et al., 2017)	-	39.87	15.82	36.90	A Deep Reinforced Model for Abstractive Summarization	-
MatchSum (BERT-base)	-	44.22	20.62	40.38	Extractive Summarization as Text Matching	-
BigBird-Pegasus	-	43.84	21.11	40.74	Big Bird: Transformers for Longer Sequences	-
DynamicConv	-	39.84	16.25	36.73	Pay Less Attention with Lightweight and Dynamic Convolutions	-
BertSumExt	-	43.85	20.34	39.9	Text Summarization with Pretrained Encoders	-
HAT-BART	-	44.48	21.31	41.52	Hierarchical Learning for Generation with Long Source Sequences	-
NeuSUM	-	41.59	19.01	37.98	Neural Document Summarization by Jointly Learning to Score and Select Sentences	-
GPT-2	-	29.34	8.27	26.58	Language Models are Unsupervised Multitask Learners	-
LightConv	-	39.52	15.97	36.51	Pay Less Attention with Lightweight and Dynamic Convolutions	-
Fourier Transformer	-	44.76	21.55	41.34	Fourier Transformer: Fast Long Range Modeling by Removing Sequence Redundancy with FFT Operator	-
Synthesizer (R+V)	-	38.57	16.24	35.95	Synthesizer: Rethinking Self-Attention in Transformer Models	-
Selector+Pointer Generator	-	41.72	18.74	38.79	Mixture Content Selection for Diverse Sequence Generation	-
UniLM (Abstractive Summarization)	-	43.08	20.43	40.34	Unified Language Model Pre-training for Natural Language Understanding and Generation	-
ML + Intra-Attention (Paulus et al., 2017)	-	38.30	14.81	35.49	A Deep Reinforced Model for Abstractive Summarization	-

0 of 26 row(s) selected.