Predicting the gender of Korean personal names using fastText
0
조회
0
다운로드
2
인용
0
참고문헌
키워드
초록
Male and female names tend to have distinct phonotactic characteristics in many languages. This paper explores the use of fastText, a neural-network text-classifier using sub-word information, in predicting the gender of Korean personal names, and compares the results with the results from a maximum-entropy model of phonotactics (Hayes and Wilson 2008). In this study, fastText is trained with training data consisting of 6400 Korean personal names, labeled with male and female. The model is tested with testing data of 35 Korean names. The fastText results positively correlated with Korean speakers’ ratings on the gender of the names. It outperformed the maximum-entropy model in terms of correlation with human ratings and accuracy of the labels. Yet, while the maximum-entropy model has OT-style constraints allowing generative linguists to interpret the results, fastText does not offer such interpretability. An error analysis is presented for the names where the models made incorrect predictions, using OT-style constraints.
Abstract
Male and female names tend to have distinct phonotactic characteristics in many languages. This paper explores the use of fastText, a neural-network text-classifier using sub-word information, in predicting the gender of Korean personal names, and compares the results with the results from a maximum-entropy model of phonotactics (Hayes and Wilson 2008). In this study, fastText is trained with training data consisting of 6400 Korean personal names, labeled with male and female. The model is tested with testing data of 35 Korean names. The fastText results positively correlated with Korean speakers’ ratings on the gender of the names. It outperformed the maximum-entropy model in terms of correlation with human ratings and accuracy of the labels. Yet, while the maximum-entropy model has OT-style constraints allowing generative linguists to interpret the results, fastText does not offer such interpretability. An error analysis is presented for the names where the models made incorrect predictions, using OT-style constraints.
참고문헌 (28)
- BOJANOWSKI, PIOTR, "Enriching word vectors with subword information", Transactions of the Association for Computational Linguistics, 5, pp.135~146 (2017)[DOI]
- 조혜선, "A maximum-entropy model of phonotactics for Korean male and female names", 음성음운형태론연구, 27(1), pp.99~129 (2021), 한국음운론학회[DOI]
- CUTLER, ANNE, "Elizabeth and John:sound patterns in men’s and women’s names", Journal of Linguistics, 26, pp.471~482 (1990)
- GOODFELLOW, IAN, "Deep Learning" (2016), The MIT Press
- GÜLER, GÖKHAN, "Comparison of Turkish word representations trained on different morphological forms", arXiv. 2002.05417, 107~116 (2020)
- HAYES, BRUCE., "Manual for Maxent Grammar Tool", Ms, pp.2009
- HAYES, BRUCE, "A maximum entropy model of phonotactics and phonotactic learning", Linguistic Inquiry, 39, pp.379~440 (2008)
- IONIN, TANIA, "Research Methods in Second Language Acquisition: a Practical Guide", pp.30~52 (2012), Blackwell Publishing
- EDOUARD GRAVE, "Bag of tricks for efficient text classification", Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics, 2, pp.427~431 (2017)
- KAWAHARA, SHIGETO, "Japanese loanword devoicing revisited : a rating study", Natural Language and Linguistic Theory, 29(3), pp.705~723 (2011)
- KAWAHARA, SHIGETO, "Comparing a force-choice wug test and a naturalness rating test : an exploration using rendaku", Language Sciences, 48, pp.42~47 (2015)
- KAWAHARA, SHIGETO, "Accounting for the stochastic nature of sound symbolism using Maximum Entropy model", Open Linguistics, 5, pp.109~120 (2019)
- , 1994
- 이진, "딥러닝 언어모델의한국어 학습자 말뭉치 원어민성 판단 결과 분석 연구", 언어와 문화, 17(1), pp.155~177 (2021), 한국언어문화교육학회[DOI]
- LINZEN, TAL, "What can linguistics and deep learning contribute to each other? Response to Pater", Language, 95(1), pp.e99~e108 (2019)
- PARK, KWONSIK, "A Comparison of Nativeness Judgments by the Deep Learning Systems and Humans on Korean EFL Learner Sentences", MA, pp.2020
- 박권식, "Not Yet as Native as Native Speakers: Comparing Deep Learning Predictions and Human Judgments", 영어학연구, 26(1), pp.199~228 (2020), 한국영어학학회[DOI]
- PATER, JOE, "Generative linguistics and neural networks at 60 : Foundation, friction, and fusion", Language, 95(1), pp.e41~e74 (2019)
- PRESTON, CAROLYN C., "Optimal number of response categories in rating scales : reliability, validity, discriminating power, and respondent preferences", Acta Psychologica, 104, pp.1~15 (2000)
- RAMACHANDRAN, VILAYANUR S., "Synaesthesia – a window into perception, thought and language", Journal of Consciousness Studies, 8(12), pp.3~34 (2001)
- SHINOHARA, KAZUKO, "The sound symbolic nature of Japanese maid names", Proceedings of the 13th Annual Meeting of the Japanese Cognitive Linguistics Association, pp.183~193 (2013)
- SIDHU, DAVID M., "What’s in a Name? Sound Symbolism and Gender in First Names", PLoS ONE, 10(5), pp.e0126809~ (2015)[DOI]
- SLEPIAN, MICHAEL L., "The Voiced Pronunciation of Initial Phonemes Predicts the Gender of Names", Journal of Personality and Social Psychology, 110(4), pp.509~527 (2016)
- SULLIVAN, LISA, "Phonology of Gender in French and English Given Names", MA, pp.2018
- SULLIVAN, LISA, "Phonology of gender in English and French given names", Proceedings of the 19th International Congress of Phonetic Sciences, pp.2124~2128 (2019)
- WONG, KRISTENU, "Sound symbolism of gender in Cantonese First Names", Proceedings of the 19th International Congress of Phonetic Sciences, pp.2129~2133 (2019)
- WRIGHT, SAUNDRA, "Gendered Practices in Language", pp.175~191 (2002), CSLI Publications
- WRIGHT, SAUNDRA, "Ladies first? Phonology, frequency, and the naming conspiracy", Linguistics, 43(3), pp.531~561 (2005)[DOI]