# Sentence-Level Content Planning and Style Specification for Neural Text Generation

arXiv: https://arxiv.org/abs/1909.00734 · HTML: https://mlsanity.com/paper/1909.00734/

## Key facts

- Cited by 23 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 2 Sep 2019.

## Built on by (most cited first, 23 of 23)

- [A Token-level Reference-free Hallucination Detection Benchmark for Free-form Text Generation](https://mlsanity.com/paper/2104.08704/) (2021)
- [Recent Advances in Deep Learning Based Dialogue Systems: A Systematic Survey](https://mlsanity.com/paper/2105.04387/) (2021)
- [MReD: A Meta-Review Dataset for Structure-Controllable Text Generation](https://mlsanity.com/paper/2110.07474/) (2021)
- [Argument Undermining: Counter-Argument Generation by Attacking Weak Premises](https://mlsanity.com/paper/2105.11752/) (2021)
- [NLP for Counterspeech against Hate: A Survey and How-To Guide](https://mlsanity.com/paper/2403.20103/) (2024)
- [Scientia Potentia Est -- On the Role of Knowledge in Computational Argumentation](https://mlsanity.com/paper/2107.00281/) (2021)
- [Efficient Few-Shot Fine-Tuning for Opinion Summarization](https://mlsanity.com/paper/2205.02170/) (2022)
- [Dialogue Planning via Brownian Bridge Stochastic Process for Goal-directed Proactive Dialogue](https://mlsanity.com/paper/2305.05290/) (2023)
- [Inference Time Style Control for Summarization](https://mlsanity.com/paper/2104.01724/) (2021)
- [DYPLOC: Dynamic Planning of Content Using Mixed Language Models for Text Generation](https://mlsanity.com/paper/2106.00791/) (2021)
- [Seamlessly Integrating Factual Information and Social Content with Persuasive Dialogue](https://mlsanity.com/paper/2203.07657/) (2022)
- [Knowledge-based Review Generation by Coherence Enhanced Text Planning](https://mlsanity.com/paper/2105.03815/) (2021)
- [Expository Text Generation: Imitate, Retrieve, Paraphrase](https://mlsanity.com/paper/2305.03276/) (2023)
- [UCEpic: Unifying Aspect Planning and Lexical Constraints for Generating Explanations in Recommendation](https://mlsanity.com/paper/2209.13885/) (2022)
- [Target-constrained Bidirectional Planning for Generation of Target-oriented Proactive Dialogue](https://mlsanity.com/paper/2403.06063/) (2024)
- [Controllable Dialogue Generation with Disentangled Multi-grained Style Specification and Attribute Consistency Reward](https://mlsanity.com/paper/2109.06717/) (2021)
- [Sentence-Permuted Paragraph Generation](https://mlsanity.com/paper/2104.07228/) (2021)
- [RSTGen: Imbuing Fine-Grained Interpretable Control into Long-FormText Generators](https://mlsanity.com/paper/2205.12590/) (2022)
- [ArgSciChat: A Dataset for Argumentative Dialogues on Scientific Papers](https://mlsanity.com/paper/2202.06690/) (2022)
- [Conclusion-based Counter-Argument Generation](https://mlsanity.com/paper/2301.09911/) (2023)
- [Investigating Content Planning for Navigating Trade-offs in Knowledge-Grounded Dialogue](https://mlsanity.com/paper/2402.02077/) (2024)
- [Implementing Long Text Style Transfer with LLMs through Dual-Layered Sentence and Paragraph Structure Extraction and Mapping](https://mlsanity.com/paper/2505.07888/) (2025)
- [Parallel Hierarchical Transformer with Attention Alignment for Abstractive Multi-Document Summarization](https://mlsanity.com/paper/2208.07845/) (2022)

## Abstract (arXiv)

Building effective text generation systems requires three critical components: content selection, text planning, and surface realization, and traditionally they are tackled as separate problems. Recent all-in-one style neural generation models have made impressive progress, yet they often produce outputs that are incoherent and unfaithful to the input. To address these issues, we present an end-to-end trained two-step generation model, where a sentence-level content planner first decides on the keyphrases to cover as well as a desired language style, followed by a surface realization decoder that generates relevant and coherent text. For experiments, we consider three tasks from domains with diverse topics and varying language styles: persuasive argument construction from Reddit, paragraph generation for normal and simple versions of Wikipedia, and abstract generation for scientific articles. Automatic evaluation shows that our system can significantly outperform competitive comparisons. Human judges further rate our system generated text as more fluent and correct, compared to the generations by its variants that do not consider language style.