English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

State-of-the-art poetry generation systems are often complex. They eitherconsist of task-specific model pipelines, incorporate prior knowledge in theform of manually created constraints, or both. In contrast, end-to-end modelswould not suffer from the overhead of having to model prior knowledge and couldlearn the nuances of poetry from data alone, reducing the degree of humansupervision required. In this work, we investigate end-to-end poetry generationconditioned on styles such as rhyme, meter, and alliteration. We identify andaddress lack of training data and mismatching tokenization algorithms aspossible limitations of past attempts. In particular, we successfully pre-trainByGPT5, a new token-free decoder-only language model, and fine-tune it on alarge custom corpus of English and German quatrains annotated with our styles.We show that ByGPT5 outperforms other models such as mT5, ByT5, GPT-2 andChatGPT, while also being more parameter efficient and performing favorablycompared to humans. In addition, we analyze its runtime performance anddemonstrate that it is not prone to memorization. We make our code, models, anddatasets publicly available.

ByGPT5: End-to-End Style-conditioned Poetry Generation with Token-free Language Models