Reproduce “Augmenting Poetry Composition with Verse by Verse” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
David Uthus, Maria Voitovich, R.J. Mical
Verse by Verse from Google Research is an interactive AI system providing line-by-line suggestions for poetry composition, leveraging a multi-model architecture and robust rhyme handling. Human evaluators perceived 47.0% of the AI-generated poems as human-written, demonstrating its capacity to assist creative writing with stylistic adherence.
Key insights
- Decoupling the generation of individual verses from the recommendation of contextually appropriate next verses enables an efficient, low-latency interactive system.
- The adoption of the Kestrel text normalization system is crucial for reliably processing diverse and potentially irregular user-provided text, a common challenge in interactive creative tools.
- Generating a large, filtered index of verses offline allows for thorough quality control and ethical content screening without compromising real-time performance during user interaction.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.