Neural language models are gaining popularity in real-life creative tasks such as text-adventure games, collaborative slogan writing, and even sports journalism, poetry and novel generation. Most such language models however provide limited interaction support for users, as control that goes beyond simple left-to-right text generation requires explicit training.
To address this limitation, a team from Google Research has proposed Wordcraft, a text editor with a built-in AI-powered creative writing assistant. Wordcraft leverages few-shot learning and the natural affordances of conversation to support a variety of user interactions; and can help with story planning, writing and editing.
The Wordcraft web interface comprises a traditional text editor augmented with a number of key commands for triggering requests to the AI assistant. The model is able to sketch a story outline, write the story and even perform editing and rewrites. For example, the command “get continuations” will generate additional context-based text from the position of the user’s cursor, while the “rewrite” and “elaborate” commands can be used to rephrase or elaborate on selected text blocks. Users can also ask Wordcraft to perform specific tasks such as “Help me describe the elderly man’s emotional state,” and try out their own queries by modifying the prompts.
To support their novel human-AI writing assistant collaboration system, the team employed Meena (Adiwardana et al., 2020), an open-ended dialogue system for text generation. Originally designed as a chatbot, Meena is broadly capable of following instructions and answering questions posed in a conversational format. The researchers also applied few-shot learning to enable their language model to perform desired tasks based on users’ instructions. Moreover, if the AI assistant does not understand what a human writer is requesting, it will return a request for clarification in a natural language conversational format, making it user-friendly.
Some of the interactions built into Wordcraft include continuation, infilling, elaboration and rewriting, and the paper provides extensive examples for each. The stories generated via Wordcraft’s human-AI collaboration editor are natural and interpretable, demonstrating the promise and potential for such human-AI collaboration language models.
The researchers regard Wordcraft as a starting point for further studies on human collaboration with AI-powered creative writing assistants. Their future plans include conducting more formal user studies to improve understanding of what assistance writers require and how Wordcraft’s natural language generation can be expanded to meet these needs.
The paper Wordcraft: A Human-AI Collaborative Editor for Story Writing is on arXiv.
Author: Hecate He | Editor: Michael Sarazen, Chain Zhang
We know you don’t want to miss any news or research breakthroughs. Subscribe to our popular newsletter Synced Global AI Weekly to get weekly AI updates.