ALBERTI

ALBERTI is a set of two BERT-based multilingual model for poetry. One for verses and another one for stanzas. This model has been further trained with the PULPO corpus for verses using Flax, including training scripts.

This is part of the Flax/Jax Community Week, organised by HuggingFace and TPU usage sponsored by Google.

PULPO

PULPO, the Prodigious Unannotated Literary Poetry Corpus, is a set of multilingual corpora of verses and stanzas with over 95M words.

The following corpora has been downloaded using the Averell tool, developed by the POSTDATA team:

Spanish

• Disco v3
• Corpus of Spanish Golden-Age Sonnets
• Corpus general de poesía lírica castellana del Siglo de Oro
• Gongocorpus - source

English

• Eighteenth-Century Poetry Archive (ECPA)
• For better for verse

French

• Métrique en Ligne - source

Italian

• Biblioteca italiana - source

Czech

• Corpus of Czech Verse

Portuguese

• Stichotheque

Also, we obtained the following corpora from these sources:

Spanish

• Poesi.as - source

English

• A Gutenberg Poetry Corpus

Arabic

• Arabic Poetry dataset

Chinese

• THU Chinese Classical Poetry Corpus

Finnish

• SKVR

German

• TextGrid Poetry Corpus - source
• German Rhyme Corpus

Hungarian

• verskorpusz

Portuguese

• Poems in Portuguese

Russian

• 19 000 Russian poems

Team members

• Álvaro Pérez (alvp)
• Javier de la Rosa (versae)
• Aitor Díaz (aitordiaz)
• Elena González-Blanco
• Salvador Ros (salva)

Useful links

• Community Week timeline
• Community Week README
• Community Week thread
• Community Week channel
• Masked Language Modelling example scripts
• Model Repository

Acknowledgments

This project would not have been possible without the infrastructure and resources provided by HuggingFace and Google Cloud. Moreover, we want to thank POSTDATA Project (ERC-StG-679528) and the Computational Literary Studies Infrastructure (CLS INFRA No. 101004984) of the European Union's Horizon 2020 research and innovation programme for their support and time allowance.