Style Control for Schema-Guided Natural Language Generation

TL;DR

This paper explores methods for controlling style in schema-guided natural language generation, balancing semantic accuracy and stylistic attributes using various training and decoding techniques, evaluated through automatic and human metrics.

Contribution

It compares controlled generation methods like conditional training, guided fine-tuning, and guided decoding for style control in schema-guided NLG, highlighting their advantages and limitations.

Findings

Conditional training works well for lexically-defined styles.

Guided decoding achieves control for complex styles.

Scalable methods with disentangled content and style are most effective.

Abstract

Natural Language Generation (NLG) for task-oriented dialogue systems focuses on communicating specific content accurately, fluently, and coherently. While these attributes are crucial for a successful dialogue, it is also desirable to simultaneously accomplish specific stylistic goals, such as response length, point-of-view, descriptiveness, sentiment, formality, and empathy. In this work, we focus on stylistic control and evaluation for schema-guided NLG, with joint goals of achieving both semantic and stylistic control. We experiment in detail with various controlled generation methods for large pretrained language models: specifically, conditional training, guided fine-tuning, and guided decoding. We discuss their advantages and limitations, and evaluate them with a broad range of automatic and human evaluation metrics. Our results show that while high style accuracy and semantic correctness are easier to achieve for more lexically-defined styles with conditional training, stylistic control is also achievable for more semantically complex styles using discriminator-based guided decoding methods. The results also suggest that methods that are more scalable (with less hyper-parameters tuning) and that disentangle content generation and stylistic variations are more effective at achieving semantic correctness and style accuracy.