Sentence Simplification with Deep Reinforcement Learning
- Code: https://github.com/XingxingZhang/dress
- 584 | applications as preprocessing step for other NLP tasks, talks about target-groups-for-text-simplification
- 585 | neural encoder-decoder RNN framework with RL
- Newsela and Wikipedia
- 586 | dataset targets all have large portions of directly copied text (extraction)
- 586 | Reward algorithm is based on simplicity (SARI in both directions), relevance (meaning-preservation through cosine similarity of input and output sequence vector) and fluency
Metadaten ( PDF)
- Zusammenfassung:: Using Reinforcement Learning on the simplification problem, Zhang et al. show that using neural models leads to better performance on common benchmarks such as BLEU and SARI.
- Motivation:: First RL study on simplification
- Ergebnisse::
- Keywords: text-simplification
- Zotero öffnen
Highlights
Imported: 2023-03-16 14:49
⭐ Main ideas
- “Our model, which we call DRESS (as shorthand for Deep REinforcement Sentence Simplification), explores the space of possible simplifications while learning to optimize a reward function that encourages outputs which are simple, fluent, and preserve the meaning of the input.” (p. 584)
✅ Useful
- “The main goal of sentence simplification is to reduce the linguistic complexity of text, while still retaining its original information and meaning.” (p. 584)
🧩 Methodology
- “it explores the space of possible simplifications while learning to maximize an expected reward function that encourages outputs which meet simplificationspecific constraints” (p. 585)
- “The reward r(ˆY) for system output ˆY is the weighted sum of the three components aimed at capturing key aspects of the target output, namely simplicity, relevance, and fluency” (p. 586)
📚 Investigate
- “Advaith Siddharthan. 2004. Syntactic simplification and text cohesion. in research on language and computation. Research on Language and Computation, 4(1):77–109.
- Advaith Siddharthan. 2014. A survey of research on text simplification. International Journal of Applied Linguistics, 165(2):259–298.” (p. 594)