UnMASKed: Quantifying Gender Biases in Masked Language Models through Linguistically Informed Job Market Prompts
Fuente:
arXiv
Salvato in:
| Autore principale: | |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866916108831817728 |
|---|---|
| author | Parra, Iñigo |
| author_facet | Parra, Iñigo |
| contents | Language models (LMs) have become pivotal in the realm of technological advancements. While their capabilities are vast and transformative, they often include societal biases encoded in the human-produced datasets used for their training. This research delves into the inherent biases present in masked language models (MLMs), with a specific focus on gender biases. This study evaluated six prominent models: BERT, RoBERTa, DistilBERT, BERT-multilingual, XLM-RoBERTa, and DistilBERT-multilingual. The methodology employed a novel dataset, bifurcated into two subsets: one containing prompts that encouraged models to generate subject pronouns in English, and the other requiring models to return the probabilities of verbs, adverbs, and adjectives linked to the prompts' gender pronouns. The analysis reveals stereotypical gender alignment of all models, with multilingual variants showing comparatively reduced biases. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2401_15798 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | UnMASKed: Quantifying Gender Biases in Masked Language Models through Linguistically Informed Job Market Prompts Parra, Iñigo Computation and Language Language models (LMs) have become pivotal in the realm of technological advancements. While their capabilities are vast and transformative, they often include societal biases encoded in the human-produced datasets used for their training. This research delves into the inherent biases present in masked language models (MLMs), with a specific focus on gender biases. This study evaluated six prominent models: BERT, RoBERTa, DistilBERT, BERT-multilingual, XLM-RoBERTa, and DistilBERT-multilingual. The methodology employed a novel dataset, bifurcated into two subsets: one containing prompts that encouraged models to generate subject pronouns in English, and the other requiring models to return the probabilities of verbs, adverbs, and adjectives linked to the prompts' gender pronouns. The analysis reveals stereotypical gender alignment of all models, with multilingual variants showing comparatively reduced biases. |
| title | UnMASKed: Quantifying Gender Biases in Masked Language Models through Linguistically Informed Job Market Prompts |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2401.15798 |