arXiv Research: Breaking and Permeability of Gender Binary in Science Fiction Language
Summary:arXiv has released a study analyzing the linguistic representation of gender in science fiction and general fiction corpora. By masking terms like 'man/men', 'woman/women', and 'person/people' and classifying RoBERTa's top five substitutions by gender, the research reveals that science fiction exhibits greater gender permeability and a stronger tendency toward degendering compared to general fiction. Specifically, the model predicts masculine terms for masked 'woman' about twice as often as it p
Background and Motivation
Science fiction is often credited with challenging and subverting traditional gender norms. This study investigates whether this phenomenon is reflected at the linguistic level, particularly in how gendered references (e.g., 'man/men', 'woman/women', 'person/people') are substituted in science fiction compared to general fiction.
Methodology
The study followed these steps:
- Data Processing: Corpora of science fiction and general fiction from the same time period were collected, and target terms were masked.
- Model Prediction: RoBERTa was used to predict the top five likely substitutions for the masked terms, which were then classified by gender.
- Statistical Analysis: Permutation tests over five resampling rounds analyzed patterns of retention and substitution, and the underlying sentences were closely read to verify the results.
Key Findings
-
Breaking the Gender Binary: Both corpora showed a movement away from gendered terms toward their counterparts, neutral terms, or other social categories, but the patterns were uneven. For instance, the model predicted masculine terms for masked 'woman' about twice as often as it predicted feminine terms for masked 'man'.
-
Uniqueness of Science Fiction: Science fiction exhibited greater gender permeability and a stronger tendency toward degendering compared to general fiction. It more frequently placed male references in feminine contexts and showed a more pronounced shift toward gender-neutral language.
-
Substitution with Social Categories and Gender-Neutral Terms: The model predicted social category terms and gender-neutral terms for masked 'woman' far more often than for masked 'man'.
Engineering Trade-offs and Practical Applications
- Model Applicability: While RoBERTa performed well in handling complex language patterns, its predictions still require human interpretation to avoid overgeneralization.
- Cross-Domain Applications: The research methodology can be extended to analyze other social category vocabularies, such as race and class, providing a new tool for computational literary studies.
- Data Requirements: The study relies on large-scale, high-quality corpora, which can be costly to obtain and process.
Conclusion and Future Directions
This study provides linguistic evidence for the thematic claims in science fiction studies and offers a pipeline applicable to other category vocabies in computational literary studies. Future research could further explore gender language patterns in different cultural contexts and incorporate multimodal data (e.g., images, audio) for a more comprehensive analysis.
— END —Source: ArXiv NLP/LLM (cs.CL) (2026-10-15)
Tags: #Gender Studies #Computational Literature #RoBERTa #Science Fiction #Language Models
Editorial & Fact-Checking Note: This article is compiled from primary research, official release documentation, and source papers by the ZICQ Newsroom pipeline with automated entity verification and human editorial review. If you notice any technical inaccuracy, please submit a correction via our corrections policy or email our editorial desk directly.
Community Comments