View Proposal

---

Proposer

Simona Frenda

Title

Gender in culture and language: from one language to another

Goal

Exploration of gender-fair language and exploration of model's gender bias, modelling gender fair models or new evaluation approaches to measure model fairness.

Description

Is machine translation gender fair? Gender in English is represented mainly semantically with important social implications. There are specific female and male nouns (i.e., aunt, queen, soul sister vs. uncle, king, sugar daddy) but most of the personal nouns are used to refer to both female and male referents (i.e., person, neighbor, engineer, babysitter, movie star, addict). However, the choice of anaphoric pronouns suggests the “psychological gender”, e.g., “The babysitter is doing a great work with Tom, she is a real pedagogical expert”. Generative and machine translation models can embed this semantic bias, especially when we ask them to translate from a gender-marked language (i.e., Italian, Spanish, Greek, and so on) to English. The neutralisation of gender or the use of all genders in sentences has an important impact on the listeners' minds, encouraging, for instance, young people to choose careers and educational paths that they thought were only for men because of the language. The student will explore this phenomenon from a multilingual perspective and could implement new approaches to 1. explore the model's gender bias, 2. post-train or reinforcement of existing models, making them gender fair, or 3. define new evaluation methods to measure the model's fairness.

Resources

- Hellinger, M., & Bußmann, H. (2003). Gender across languages: The linguistic representation of women and men. Vol. 1 (pp for the English language: 1-26, 105-114) - Savoldi, B., Gaido, M., Bentivogli, L., Negri, M., & Turchi, M. (2021). Gender bias in machine translation. Transactions of the Association for Computational Linguistics, 9, 845-874. - Rosola, M., Frenda, S., Cignarella, A. T., Pellegrini, M., Marra, A., & Floris, M. (2023, November). Beyond obscuration and visibility: Thoughts on the different strategies of gender-fair language in Italian. In Proceedings of the Ninth Italian Conference on Computational Linguistics (CLiC-it 2023) (pp. 369-378). - Blodgett, S. L., Lopez, G., Olteanu, A., Sim, R., & Wallach, H. (2021, August). Stereotyping Norwegian salmon: An inventory of pitfalls in fairness benchmark datasets. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) (pp. 1004-1015).

Background

Url

Difficulty Level

High

Ethical Approval

None

Number Of Students

0

Supervisor

Simona Frenda

Keywords

machine translation, gender bias, generative tasks

Degrees

Bachelor of Science in Computer Science
Master of Science in Artificial Intelligence
Master of Science in Computing (2 Years)
Master of Science in Data Science
Bachelor of Science in Computing Science
Postgraduate Diploma in Artificial Intelligence
Bachelor of Science in Statistical Data Science
BSc Data Sciences