Isabelle Zaugg, Smaranda Muresan Mentor Finalists in Data Science Student Course Projects Competition
March 28, 2022 – Achievements
Smaranda Muresan and Isabelle Zaugg’s Multilingual Technologies and Language Diversity class, funded by a collaborative grant through the Data Science Initiative, produced two successful finalists in the 2022 Best Data Science Student Course Projects Finalist Competition, organized by their students Tuhin Hakrabarty, Arkadiy Saakyan, Eve Suane Loomis Washington, Gabriela Arredondo, Megan Fleurine St. Hilaire, and David Rosado. For more information about the program, visit the link here.
Don’t Go Far Off: An Empirical Study on Neural Poetry Translation
- Team Members: Tuhin Hakrabarty, Arkadiy Saakyan
- Instructor: Smaranda Muresan
- Short Description: Despite constant improvements in machine translation quality, automatic poetry translation remains a challenging problem due to the lack of open-sourced parallel poetic corpora, and to the intrinsic complexities involved in preserving the semantics, style and figurative nature of poetry. We present an empirical investigation for poetry translation along several dimensions: 1) size and style of training data (poetic vs. non-poetic), including a zero-shot setup; 2) bilingual vs. multilingual learning; and 3) language-family-specific models vs. mixed-language-family models. To accomplish this, we contribute a parallel dataset of poetry translations for several language pairs. Our results show that multilingual fine-tuning on poetic text significantly outperforms multilingual fine-tuning on non-poetic text that is 35X larger in size, both in terms of automatic metrics (BLEU, BERTScore, COMET) and human evaluation metrics such as faithfulness (meaning and poetic style). Moreover, multilingual fine-tuning on poetic data outperforms bilingual fine-tuning on poetic data.
Script Key: An Image-Based Keyboard for Non-Encoded Alphabetic Scripts
- Team Members: Eve Suane Loomis Washington, Gabriela Arredondo, Megan Fleurine St. Hilaire, David Rosado
- Instructor: Isabelle Zaugg
- Short Description: This project sought to develop a tool that would enable users to send messages using unencoded scripts. Script Key allows for generating a library of symbol images, not only for use in messages, but also for crowdsourcing examples of a script’s use that are required for Unicode standardization.
