Read, Revise, Repeat: A System Demonstration for Human-in-the-loop Iterative Text Revision

Abstract

Revision is an essential part of the human writing process. It tends to be strategic, adaptive, and, more importantly, iterative in nature. Despite the success of large language models on text revision tasks, they are limited to non-iterative, one-shot revisions. Examining and evaluating the capability of large language models for making continuous revisions and collaborating with human writers is a critical step towards building effective writing assistants. In this work, we present a human-inthe-loop iterative text revision system, Read, Revise, Repeat (R3), which aims at achieving high quality text revisions with minimal human efforts by reading model-generated revisions and user feedbacks, revising documents, and repeating human-machine interactions. In R3, a text revision model provides text editing suggestions for human writers, who can accept or reject the suggested edits. The accepted edits are then incorporated into the model for the next iteration of document revision. Writers can therefore revise documents iteratively by interacting with the system and simply accepting/rejecting its suggested edits until the text revision model stops making further revisions or reaches a predefined maximum number of revisions. Empirical experiments show that R3 can generate revisions with comparable acceptance rate to human writers at early revision depths, and the human-machine interaction can get higher quality revisions with fewer iterations and edits. The collected human-model interaction dataset and system code are available at https://github. com/vipulraheja/IteraTeR. Our system demonstration is available at https:// youtu.be/lK08tIpEoaE.

A Design Space for Intelligent and Interactive Writing Assistants

“…With deep sequence to sequence models [42,239] writing assistants were able to use a combination of human-labeled and machine-labeled data with several thousand sentences or examples [160,162]. Several recent works [35,45,58,65,174,188,268] take advantage of the large language models that are already pre-trained on corpora of millions of sentences. They are then able to use small amounts of human-labeled data to tune the model (e.g., instruction tune [204,218,226]) and in many cases simply prompt the model in a zero-shot manner to help with the writing task.…”

Exploring Sentence-level Revision Capabilities of LLMs in English for Academic Purposes Writing Assistance

“…Another task, SentRev, serves as an enhanced version of GEC, aiming not only to address GEC but also to polish academic English style and improve text fluency. However, despite various attempts [5,6] to advance this task from multiple perspectives, the outcomes have been less than ideal.…”

A Survey of Reasoning with Foundation Models

“…The Learned Refiner necessitates a training process, and the acquisition of supervised refinement typically involves pairs of feedback and refinement (Schick et al, 2022;Du et al, 2022b;Yasunaga and Liang, 2020;Madaan et al, 2021). CURIOUS (Madaan et al, 2021) initially constructs a graph that represents relevant influences.…”

“…PEER (Schick et al, 2022) is an advanced collaborative language model that replicates the entire writing process, encompassing drafting, suggesting modifications, proposing edits, and providing explanations for its actions. In contrast, Read, Revise, Repeat (R3) (Du et al, 2022b) aims to achieve superior text revisions with minimal human intervention. It achieves this by analyzing model-generated revisions and user feedback, making document revisions, and engaging in repeated human-machine interactions.…”