This project explores the complexity of inferring gender from Chinese names and the limitations of using Pinyin for that purpose. While Pinyin simplifies Chinese characters for global audiences, it compresses thousands of characters with distinct meanings, tones, and cultural associations into a small set of shared syllables. As a result, many names that are clearly gendered in Chinese become ambiguous when written in Pinyin. Our aim is to illustrate how gender cues are embedded in Chinese characters, why these cues are lost in romanization, and how this affects perception and prediction of gender in practice.
We use Hanyu Pinyin, the standard romanization for Mandarin since 1958, recognized internationally in 1982 and by Taiwan in 2009. However, many prominent Chinese figures—especially those known internationally before Pinyin’s adoption—retain names in older systems like Wade-Giles. For example, Chen-Ning Yang (杨振宁) follows Wade-Giles conventions, which differ from the Pinyin form Yang Zhenning.
To explore more about name ambiguity created by Pinyin transliteration, see the related project Not My Name. To learn more about the broader research and design initiative, see Lost in Transliteration: Chinese Names to the West.
This project analyzes gender patterns in commonly used Chinese characters found in given names. Key methods and considerations include:
Character Selection: Starting with 8,105 General Standard Chinese Characters from HanziDB, we identified their Pinyin pronunciations and meanings from Hanzi Quanxi. Neutral tones and characters that only have a neutral tone were excluded, as names typically use one of four standard tones. Pronunciations without a recorded meaning in Hanzi Quanxi were also removed.
Meaning Translation: Character definitions were translated into English using DeepSeek AI, with prompts directing it to provide conceptual, culturally accessible meanings rather than technical or dictionary-literal translations.
Gender Analysis: Using data from the Chinese Name Database 1930-2008, we analyzed characters used in given names, selecting those with a Name-Character Uniqueness (NU) > 6 and a valid Name-Gender (NG) value, resulting in 2,603 characters.
Pinyin-level gender was estimated by the NU-weighted average NG of all characters sharing the same syllable. The reported 24% gender-pattern retention corresponds to the proportion of variance explained (r²) between character-level NG and syllable-level weighted NG.
Gender labels were assigned to characters based on their gender score: strongly female, slightly female, neutral, slightly male, and strongly male, using thresholds of -0.6, -0.2, 0.2, and 0.6.
Exclusion of Complex Names: We excluded given names longer than three characters. Three characters are shown only when they cannot be split into one or two characters.
Handling Multiple Pronunciations: For characters with multiple pronunciations, we prioritize the pronunciation typically used in given names. However, we retain other pronunciations as there are no strict rules preventing alternate pronunciations from being used in names.
Audio pronunciations are sourced from Chinese Pronunciation Wiki (©2015-2022 AllSet Learning).