The First Workshop on Intelligent and Interactive Writing Assistants

The purpose of this interdisciplinary workshop is to bring together researchers from the natural language processing (NLP) and human-computer interaction (HCI) communities as well as industry practitioners and professional writers to discuss innovations in building, improving, and evaluating intelligent and interactive writing assistants. We plan to alternate our workshop venue between an NLP conference and a HCI conference every year to facilitate collaboration.

The first 100 participants get a free premium subscription to Grammarly and Wordtune.

This year the workshop will be held at ACL 2022 in Dublin, Ireland on the 26th of May, 2022.

The workshop is in a remote/in-person hybrid format. Join zoom for the workshop (please turn on video, but mute audio upon entry)!

~*~ Accepted Papers ~*~

Invited Speakers


Timezone: Irish Standard Time - GMT+1
Location: Liffey Meeting Room 1 (Floor 1)

🗣️ - speaker will be in person (broadcast to zoom)
💻 - speaker will be joining remotely (broadcast to Dublin)

Unless otherwise stated, everything will be broadcast via zoom.

Time Event
09:00-09:10AM 🗣️ Opening remarks - Mina Lee (organizer)
Introduction to workshop. Overview of goal, submitted papers, and acknowledgements.
09:10-10:10 🗣️ Invited talk - Claire L. Evans (professional writer)
How to Chain Trip
In 2018 the pop group YACHT went into the studio to record their seventh album, Chain Tripping. Seeking a challenge, they'd decided to investigate machine learning as a compositional tool, creating their own method at the intersection of DIY and high-tech. The lyrics for Chain Tripping, created in collaboration with the creative technologist Ross Goodwin, emerged from a character-recurrent neural network trained on a corpus of about two million song lyrics. This model generated massive amounts of lyrical data at a range of temperatures—a formless abundance of raw material for the creation of new songs. In her talk “How To Chain Trip,” YACHT’s Claire L. Evans explains the band’s approach to this unconventional material, and how she—as a songwriter—drew inspiration from Dadaist poetry and David Bowie to create ear-worms from this indeterminacy.
10:10-10:30 💻 Invited talk - Daniel Gissin (industry)
Wordtune - An AI Writing Companion
The space of AI writing assistants today can be split into three main clusters - assistants that correct, assistants that write and assistants that rewrite. Wordtune falls mostly under the third category, leaning on paraphrasing as the main driver for writing improvement. In this talk, I’ll describe Wordtune and explain our rationale behind focusing on rewriting, which helps writers improve their writing while keeping them at the driver’s seat.
10:30-11:00 Coffee break
in person - make some friends
remote - themed breakout rooms on: 1) artistic pursuits; 2) professionalized contexts; and 3) general purpose v. custom tools
11:00-11:15 🗣️ Invited talk - Melissa Roemelle (academia)
Tell me without telling me: evaluating AI writing assistants from user interaction
A conventional approach to evaluating text generation within writing assistance applications is to ask users: “How helpful was this text?" But we can also measure helpfulness implicitly from users’ interaction with the generated text without this explicit feedback. I’ll describe how I’ve applied this analysis in my research and how it provides particular insight into developing more helpful writing assistants.
11:15-11:30 🗣️ Invited talk - Elizabeth Clark (academia)
Human-Machine Collaborative Writing for Better NLG Evaluation
Human evaluations of generated text are increasingly difficult to collect, in part because of the improving fluency of natural language generation (NLG) models. While a typical human evaluation task asks evaluators to rate a static set of output, human-machine collaborative writing systems allow people to directly interact with generated text. In this talk, I will demonstrate how a collaborative writing system can be used to evaluate NLG models in a downstream task. I will also discuss challenges in NLG evaluation today and how evaluation through collaborative writing addresses some of these challenges.
11:30-12:30PM 💻 Panel discussion - Understanding the impact of writing assistants on ownership, authenticity, originality, & confidence
Jill Burstein, Dashiel Carrera, Ekaterina Kochmar, and Thiemo Wambsganss.
12:30-02:00 Lunch break
02:00-03:00 🗣️ Invited talk - Lillian-Yvonne Bertram (professional writer)
Ars Combinatoria: A Poetics of Computation
While the majority of all writing today can be considered "digital" in some way, given the predominance of writing and composition on computers and other machines, the practice of writing with computation and generating creative texts spans at least the past 60 years. Early computer-generated writing experiments included genres like chatbots, poems, and letters. In this talk, poet Lillian-Yvonne Bertram will discuss their introduction to using computation for poetry and how computation, machine learning, and other aspects of NLP play a role in their writing and composition practice.
03:00-03:30 Coffee break
in person - make some friends
remote - themed breakout rooms on: 1) creativity; 2) authenticity; and 3) evaluation
03:30-04:00 🗣️ Invited talk - Timo Mertens (industry)
Towards Super-human Communication Assistants
Communication is a fundamental human need that drives meaningful connection and results—but it’s hard to get right. Even basic grammar and spelling are challenging, and while the English language is heavily rule-based, there are a lot of grey areas. Getting only the foundations of writing right is no longer an option as our remote-first reality commands more asynchronous communication. This talk will explore how an human-in-the-loop approach to Machine Learning can empower communication assistance to overcome complex communication challenges.
04:00-04:15 💻 Best paper talk - Wanyu Du and Zae Myung Kim (academia)
Read, Revise, Repeat: A System Demonstration for Human-in-the-loop Iterative Text Revision
In this work, the authors present a human-in-the-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.
04:15-05:15 🗣️/💻 Panel discussion - Bridging NLP and HCI to design, build, and evaluate writing assistants
Courtney Napoles, Melissa Roemmele, Qian Yang, and Sherry Wu
05:15-05:17 🗣️ Closing remarks - Vipul Raheja (organizer)
05:17-06:00 Demo and poster session
in person - normal poster session (location: Forum)



You can contact the organizers by emailing

Program Committee