Elamur Research⌕
☰
home / papers / 2025 / Remedying Gender Bias in Open English Wordnet

Remedying Gender Bias in Open English Wordnet

Paper · 2025 · John P. McCrae, Haotian Zhu, Fei Xia et al.

Summary

Open English Wordnet (OEWN), the open-source continuation of Princeton WordNet (PWN), carries gender biases from the time it was built. The authors adopt a taxonomy combining the work of Doughman et al. (2021) with that of Hitti et al. (2019), and they extend it with two categories of their own: distributional bias and definitional bias. They then repair the resource. They collect gendered role words, such as "-man"/"-woman" compounds and Wiktionary's list of female-person terms, and filter out false positives by hand. Gendered synsets are then attached to explicit male or female hypernyms, and matching gender-neutral or other-gender synsets are created where they were missing. Definitions that used generic "he" are rewritten with singular "they". In total, 317 synsets were added and more than 400 others had their definitions or relations revised, all shipped in OEWN 2024. A comparison of definition lengths shows that female role definitions in OEWN 2024 are less short than in PWN. The imbalance of named people, however, remains large. This matters because NLP tools that rely on the wordnet should not be pushed toward male senses when the context is clearly female.

Key points

  • Five bias types are used: explicit marking of sex (B1), generic he (B2), contextual bias (B3), and the newly added distributional (B4) and definitional (B5) bias.
  • Candidate gendered words came from two sources. One was a Wiktionary category. The other was words descending from the 'man' [10306910-n] and 'woman' [10807146-n] synsets in OEWN 2023. Of these, 556 out of 1092 turned out to be truly gendered; words like "German" or "brahman" were excluded.
  • Gendered synsets now sit under 'male' [09647338-n] or 'female' [09642198-n].
  • Female role definitions became longer between PWN 3.1 and OEWN 2024:
    • average word count rose from 8.08 to 8.67;
    • average character count rose from 42.83 to 46.83;
    • male and neutral definitions stayed about the same.
  • Gender for named people was taken from a partial Wikidata mapping (property P21). OEWN 2024 has 6,778 male name lemmas versus 845 female ones, and a moratorium on proper nouns prevents fixing this gap.

Interesting ideas

  • I1. Gender is encoded as a hypernym link to explicit 'male'/'female' synsets rather than as a loose attribute, and every role gets a neutral synset alongside its gendered ones; this keeps the property queryable through ordinary taxonomy traversal.
  • I2. Candidate lists are harvested in two ways, from a curated external list and from a morphological pattern restricted to one taxonomic subtree, and then checked by a human to remove look-alikes.
  • I3. Definition length is treated as a cheap, measurable proxy for how carefully different groups are described, and it is compared across releases.
  • I4. Missing attributes are filled from an external knowledge base (Wikidata P21) through only a partial mapping, and the uncovered cases are reported honestly.
  • I5. The paper warns against forcing disambiguation to a male sense when context indicates a woman, and adopts singular "they" as the neutral pronoun.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (world): Our WordNet-derived ontology could read entity gender from the hypernym path, so "policewoman" yields a female entity while "police officer" stays unspecified rather than defaulting to male.
  • I2 (tokens): Combining a suffix rule with a hypernym-subtree check could help the morphology layer separate real "-man" agent compounds from accidental matches like "German".
  • I3 (learn): A similar per-group statistic, such as gloss length or frame coverage per entity type, could act as a regression check on whether our compiled knowledge treats categories unevenly.
  • I4 (world): Our knowledge graph could fill person attributes from Wikidata where a link exists and mark the rest as unknown rather than guessing.
  • I5 (frames): Coreference should treat "they" as a possibly singular antecedent match and prefer a neutral role sense unless gender is explicit in the text.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓University of WashingtonSeattle · United States353 works · 200,043 citations🎓Ollscoil na Gaillimhe – University of GalwayGalway · Ireland23 works · 2,229 citations🔬Insight SFI Research Centre for Data AnalyticsDublin · Ireland9 works · 543 citations

 Fields, topics and concepts

📚 Shelves 7The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksSocial Sciences · 1,772 worksSafety Research · 543 worksGender Studies · 52 works

🧭 Topics 2Research subjects and languages the work is about.English · 2,263 worksBias, fairness & safety · 517 works

Cite

Cite as John P. McCrae, Haotian Zhu, Fei Xia, Al Waskow and Kexin Gao. 2025. Remedying Gender Bias in Open English Wordnet. Proceedings of the 13th Global Wordnet Conference.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2025.gwc-1.16

@inproceedings{mccrae2025remedying,
  title = {{Remedying Gender Bias in Open English Wordnet}},
  author = {John P. McCrae, Haotian Zhu, Fei Xia, Al Waskow and Kexin Gao},
  booktitle = {Proceedings of the 13th Global Wordnet Conference},
  year = {2025},
  doi = {10.18653/v1/2025.gwc-1.16},
  url = {https://doi.org/10.18653/v1/2025.gwc-1.16}
}

References

35 references, 10 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Mark Balhorn. 2004. The Rise of Epicene They. Journal of English Linguistics, 32(2):79–104. ↗
  2. Tolga Bolukbasi, Kai-Wei Chang, James Zou, Venkatesh Saligrama, and Adam Kalai. 2016. Man is to computer programmer as woman is to homemaker? debiasing word embeddings. ↗
  3. Francis Bond, Piek Vossen, John McCrae, and Christiane Fellbaum. 2016. CILI: the collaborative interlingual index. In Proceedings of the 8th Global WordNet Conference (GWC), pages 50–57, Bucharest, Romania. Global Wordnet Association.
  4. Shikha Bordia and Samuel R. Bowman. 2019. Identifying and reducing gender bias in word-level language models. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Student Research Workshop, pages 7–15, Minneapolis, Minnesota. Association for Computational Linguistics.
  5. Aylin Caliskan, Joanna J. Bryson, and Arvind Narayanan. 2017. Semantics derived automatically from language corpora contain human-like biases. Science, 356(6334):183–186. ↗
  6. Jenna Cryan, Shiliang Tang, Xinyi Zhang, Miriam J. Metzger, Haitao Zheng, and Ben Y. Zhao. 2020. Detecting gender stereotypes: Lexicon vs. supervised learning methods. Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems. find
  7. Mark Davies. 2010. The Corpus of Contemporary American English as the first reliable monitor corpus of English. Literary and Linguistic Computing, 25(4):447–464. ↗
  8. Emily Dinan, Angela Fan, Adina Williams, Jack Urbanek, Douwe Kiela, and Jason Weston. 2020. Queens are powerful too: Mitigating gender bias in dialogue generation.
  9. Jad Doughman and Wael Khreich. 2022. Gender bias in text: Labeled datasets and lexicons. CoRR, abs/2201.08675. ↗
  10. Jad Doughman, Wael Khreich, Maya El Gharib, Maha Wiss, and Zahraa Berjawi. 2021. Gender bias in text: Origin, taxonomy, and implications. In Proceedings of the 3rd Workshop on Gender Bias in Natural Language Processing, pages 34–44, Online. Association for Computational Linguistics. find
  11. Crystal Duran. 2024. Actor vs. actress. Backstage. find
  12. Zahra Fatemi, Chen Xing, Wenhao Liu, and Caim-ming Xiong. 2023. Improving gender fairness of pre-trained language models without catastrophic forgetting. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 1249–1262, Toronto, Canada. Association for Computational Linguistics. find
  13. Christiane Fellbaum. 2010. WordNet, pages 231–243. Springer Netherlands, Dordrecht. find
  14. Pedro A Fuertes-Olivera and Sven Tarp. 2022. Critical lexicography at work: reflections and proposals for eliminating gender bias in general dictionaries of Spanish. Lexikos, 32(2):105–132. ↗
  15. Yasmeen Hitti, Eunbee Jang, Ines Moreno, and Carolyne Pelletier. 2019. Proposed taxonomy for gender bias in text; a filtering methodology for the gender generalization subtype. In Proceedings of the First Workshop on Gender Bias in Natural Language Processing, pages 8–17, Florence, Italy. Association for Computational Linguistics. find
  16. Lucy Li, Demszky Dorottya, Bromley Patricia, and Jurafsky Dan. 2020. Content analysis of textbooks via natural language processing: Findings on gender, race, and ethnicity in texas u.s. history textbooks. AERA Open, 6. ↗
  17. Chandler May, Alex Wang, Shikha Bordia, Samuel R. Bowman, and Rachel Rudinger. 2019. On measuring social biases in sentence encoders. ArXiv, abs/1903.10561. ↗
  18. Marek Maziarz, Maciej Piasecki, Ewa Rudnicka, Stan Szpakowicz, and Paweł K˛edzia. 2016. plWordNet 3.0 – a comprehensive lexical-semantic resource. In Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 2259–2268, Osaka, Japan. find
  19. The COL-ING 2016 Organizing Committee. find
  20. John P. McCrae, Michael Wayne Goodman, Francis Bond, Alexandre Rademaker, Ewa Rudnicka, and Luis Morgado Da Costa. 2021. The GlobalWordNet formats: Updates for 2020. In Proceedings of the 11th Global Wordnet Conference, pages 91–99, University of South Africa (UNISA). Global Wordnet Association. find
  21. John P. McCrae, Alexandre Rademaker, Francis Bond, Ewa Rudnicka, and Christiane Fellbaum. 2019. find
  22. English WordNet 2019 – an open-source WordNet for English. In Proceedings of the 10th Global Wordnet Conference, pages 245–252, Wroclaw, Poland. Global Wordnet Association.
  23. John Philip McCrae, Alexandre Rademaker, Ewa Rudnicka, and Francis Bond. 2020. find
  24. English WordNet 2020: Improving and extending a WordNet for English using an open-source methodology. In Proceedings of the LREC 2020 Workshop on Multimodal Wordnets (MMW2020), pages 14–19, Marseille, France. The European Language Resources Association (ELRA). find
  25. George A. Miller. 1995. Wordnet: a lexical database for english. Commun. ACM, 38(11):39–41. ↗
  26. Carolin Müller-Spitzer. 2023. Gender stereotypes in dictionaries: the challenge of reconciling usage-based lexicography with the role of dictionaries as social agents. Lexikos, 33(2):79–94. ↗
  27. Silvia Pettini. 2021. “One is a woman, so that’s encouraging too”. the representation of social gender in “powered by Oxford” online lexicography. Lingue e Linguaggi, (44):275–295. find
  28. Beatrice Savoldi, Marco Gaido, Luisa Bentivogli, Matteo Negri, and Marco Turchi. 2021. Gender bias in machine translation. Transactions of the Association for Computational Linguistics, 9:845–874. ↗
  29. Sabine Sczesny, Magda Formanowicz, and Franziska Moser. 2016. Can gender-fair language reduce gender stereotyping and discrimination? Frontiers in Psychology, 7. ↗
  30. William Strunk and E. B. White. 1999. The Elements of style. Allyn and Bacon : Longman, Boston (USA). find
  31. Piek Vossen. 1998. Introduction to EuroWordNet, pages 1–17. Springer Netherlands, Dordrecht. find
  32. Susanne Wagner. 2003. Gender in English pronouns: Myth and reality. Ph.D. thesis, Freiburg (Breisgau), Univ., Diss., 2003. find
  33. Linda White. 1989. Feminism and lexicography: Dealing with sexist language in a bilingual dictionary. Frontiers: A Journal of Women Studies, pages 61–64. ↗
  34. Jieyu Zhao, Tianlu Wang, Mark Yatskar, Vicente Ordonez, and Kai-Wei Chang. 2018. Gender bias in coreference resolution: Evaluation and debiasing methods. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers), pages 15–20, New Orleans, Louisiana. Association for Computational Linguistics. ↗
  35. Haotian Zhu, Kexin Gao, Fei Xia, and Mari Ostendorf. 2024. Disagreeable, slovenly, honest and un-named women? investigating gender bias in English educational resources by extending existing gender bias taxonomies. In Proceedings of the 5th Workshop on Gender Bias in Natural Language Processing (GeBNLP), pages 219–236, Bangkok, Thailand. Association for Computational Linguistics.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/mccrae-2025-remedying-gender-bias-open-english/