ACL 2024 Tutorial on Automatic and Human-AI Interactive Text Generation

Yao Dou^*1

Philippe Laban^*2

Claire Gardent³

Wei Xu¹

¹Georgia Institute of Technology, ²Salesforce AI, ³CNRS/LORIA

^*Equal Contribution

We will take Q&A at the Slido link.

About this Tutorial

In this tutorial, we focus on text-to-text generation, a class of natural language generation (NLG) tasks, that takes a piece of text as input and then generates a revision that is improved according to some specific criteria (e.g., readability or linguistic styles). This includes many useful applications, such as text simplification, paraphrase generation, style transfer, etc, which are more constrained in terms of semantic consistency and targeted language styles, compared to text summarization and open-ended text completion (e.g., story). These tasks are interesting from a technical standpoint, as they require complex combinations of lexical and syntactical transformations, stylistic control, and adherence to factual knowledge, -- all at once.

With a special focus on text simplification and revision, this tutorial aims to provide an overview of the state-of-the-art NLG research from four major aspects -- Evaluation, Model Architecture, Modeling Techniques, Human-AI Collaboration. Specifically, we will cover: (1) Evaluation of LLM-generated text: from BLEU to reward models and LLM evaluators. (2) Advances in text simplification and rewriting models. (3) Innovations in decoding methods, distillation, and diffusion language models. (4) The rise of HCI+NLP+Accessibility research for real-world usable systems.

Tutorial Schedule

The tutorial will be held on Aug 11th (time are based on UTC+7 = Bangkok local time):

Time	Section	Presenter
9:00 - 9:15	Introduction [Slides]	Philippe
9:15 - 10:00	Section 1: Evaluation of LLM-generated Text [Slides]	Yao
10:00 - 10:45	Section 2: Models - (Text Simplification and Text Rewriting) [Slides]	Claire
10:45 - 11:00	Break	-
11:00 - 11:45	Section 3: Modeling Perspectives - (decoding, distillation, and diffusion LM) [Slides]	Wei
11:45 - 12:30	Section 4: from method to usable system: Human-Centered NLP [Slides]	Philippe

Slides are subject to changes.

BibTeX

@article{dou2023automatic,
    title={Automatic and Human-AI Interactive Text Generation},
    author={Dou, Yao and Laban, Philippe and Gardent, Claire and Xu, Wei},
    journal={arXiv preprint arXiv:2310.03878},
    year={2023}
}