PK PyTorchKR Papers
Papers with Code · CC BY-SA 4.0

Text Generation

440 평가 기록 256 datasets

Text Generation is the task of generating text with the goal of appearing indistinguishable to human-written text. This task is more formally known as "natural language generation" in the literature.

Text generation can be addressed with Markov processes or deep generative models like LSTMs. Recently, some of the most advanced methods for text generation include BART, GPT and other GAN-based approaches. Text generation systems are evaluated either through human ratings or automatic evaluation metrics like METEOR, ROUGE, and BLEU.

Further readings:

<span style="color:grey; opacity: 0.6">( Image credit: Adversarial Ranking for Language Generation )</span>

개요는 각 항목의 일부입니다. 위 탭에서 전체 기록을 페이지별로 탐색할 수 있습니다.

벤치마크

Papers with Code 아카이브의 데이터셋별 기록 예시입니다. 평가 조건과 지표 방향이 검증되지 않아 순위를 의미하지 않습니다.

Dataset기록된 모델 예시MetricPaperCode
HEGR
BLEU-4: 16.7METEOR: 37.8CIDEr: 14.1
논문 -
Control Prefixes (A1, T5-large)
BLEU: 67.32
논문 코드
GAP - Me,r+γ
BLEU: 66.2ROUGE: 76.36
논문 코드
S_1^R
BLEU: 68.60NIST: 8.73METEOR: 45.25ROUGE-L: 70.82
논문 코드
FactT5B
BLEU: 67.08METEOR: 48.34FactSpotter: 99.71
논문 코드
CodeTrans-MT-Large
Smoothed BLEU-4: 21.87
논문 코드
Transformer
Smoothed BLEU-4: 25.61
논문 코드
CodeTrans-MT-Base
Smoothed BLEU-4: 26.23
논문 코드
GAP - Me,r+γ
BLEU: 35.08METEOR: 27.5ROUGE: 64.28BertScore: 93.38
논문 코드
Control Prefixes (A1, A2, T5-large)
BLEU: 62.27
논문 코드
SAE+Discriminator
G-Score (BLEU, Accuracy): 74.56
논문 코드
Control Prefixes (T5-large)
BLEU (Test set): 44.15
논문 코드

데이터셋

256개
VISTWebNLGWebNLG 2.0 (Unconstrained)E2E NLG ChallengeWebNLG 2.0 (Constrained)CodeSearchNet - JavaCodeSearchNet - JavaScriptCodeSearchNet - PhpEventNarrativeWebNLG FullYelp Review Dataset (Small)Cleaned E2E NLG Challenge

논문

라이브러리

Papers with Code 리더보드 row에 연결된 구현 저장소입니다.

github.com/rupes438/CodeGen 2 논문
github.com/desire2020/RankGAN 2 논문
github.com/harvardnlp/sa-vae 2 논문
github.com/CR-Gjx/LeakGAN 1 논문
github.com/liyzcj/leakgan-py3 1 논문
github.com/nurpeiis/LeakGAN-PyTorch 1 논문
github.com/universebh/text_generation_fsa_gan 1 논문
github.com/valko073/LyricsGANs 1 논문

하위 태스크

DG Data-to-Text Generation 44 데이터셋 · 234 rows CD Code Documentation Generation 7 데이터셋 · 52 rows DG Dialogue Generation 14 데이터셋 · 24 rows TG Table-to-Text Generation 9 데이터셋 · 22 rows TS Text Style Transfer 3 데이터셋 · 11 rows MS Multi-Document Summarization 5 데이터셋 · 10 rows FT Fact-based Text Editing 2 데이터셋 · 6 rows PG Paraphrase Generation 3 데이터셋 · 5 rows SG Story Generation 5 데이터셋 · 5 rows CT Conditional Text Generation 1 데이터셋 · 4 rows SC Spelling Correction 1 데이터셋 · 4 rows HG Headline Generation 1 데이터셋 · 2 rows