"We demonstrate that word embeddings can be used as a powerful tool to quantify historical trends and social change. (..) we develop metrics based on word embeddings to characterize how gender stereotypes and attitudes toward ethnic minorities in the USA"
Interesting read - understanding language models & word embeddings. Word embeddings quantify 100 years of gender and ethnic stereotypes | Proceedings of the National Academy of Sciences https://t.co/f8OSFIzlmt
今はアレに聞けば答えてくれそうですが、言語モデルの埋め込み表現で、100年に渡るステレオタイプの変遷を定量化しようとする試みも応用としてわかりみです。これらを実務応用できるかはちょっと不明ですが。ただ面白い。 https://t.co/o5L0HysIkC
@Hopcatchy This paper (https://t.co/ChI4c8OkLi) compares gender bias in word embeddings with tests from 1977 and 1990 where people’s association of adjectives with gender is measured; not sure if anything similar has been done for prototypes, but I ran a s
@weyesb1ood Blew my mind in 2018 how AI language models are so good at ingraining racial/gender biases that you can train a model on different time periods and objectively measure how society changes its views on minorities over time via learned word assoc
Word embeddings can be used to track gender or group biases over time. For example, this article by Garg et al. (2018) showed that women were more strongly associated with occupations in the 1990s than in the 1950s: https://t.co/dO1yprKZkO #wordembeddings
RT @alethioguy: @DrDesmondPatton This work looks at racial bias against Asian people specifically, using word embeddings of names: https://…
@DrDesmondPatton This work looks at racial bias against Asian people specifically, using word embeddings of names: https://t.co/1yHFGrirTw
Word embeddings quantify 100 years of gender and ethnic stereotypes https://t.co/HMqRUvmvbN
@postdiscipline @brianchristian Check out the abstract! https://t.co/DbsKLDfsxZ
.@brianchristian notes that machine learning allows us to track big cultural changes over time. One advantage is that cultural change is not tracked in terms of only a few classic texts or key newspapers but over a huge corpus of cultural representations.
Specifically, we study word embeddings following Garg, Schiebinger, Jurafsky & Zou (2018) and find evidence that the texts in the COHA (newspaper articles, fiction, non-fiction, and magazine articles from 1820s-2010s) reflect gender bias. https://t.co
@LiNc0RE So then the machine learning model that uses the texts to learn something (like translation or search engines) also learns the biases. Interesting reads: https://t.co/8nRQvzzZt8 https://t.co/WjF9LDFols https://t.co/3TzFSHC0rk
Word embeddings quantify 100 years of gender and ethnic stereotypes https://t.co/vgoOotK0JO via @stanfordnlp
RT @asherjdm: Garg, Schiebinger, Jurafsky, & Zou (2018) - studying the association between word embedding measures of gender bias and occup…
Garg, Schiebinger, Jurafsky, & Zou (2018) - studying the association between word embedding measures of gender bias and occupation shifts. Also looks at ethnic stereotypes and their relation to occupational adoption. https://t.co/NDVcJpmknu
RT @Shevek91: https://t.co/A8zqr8mFg3 이 연구도 과거 소개한 기억이 있는데, 이 논문만 해도 참여한 연구자들 가운데 하나 빼고는 다 CS 소속이고 처음 프로시딩 형태로 올라온 곳도 arxiv였습니다. 2018년에 출판…
https://t.co/A8zqr8mFg3 이 연구도 과거 소개한 기억이 있는데, 이 논문만 해도 참여한 연구자들 가운데 하나 빼고는 다 CS 소속이고 처음 프로시딩 형태로 올라온 곳도 arxiv였습니다. 2018년에 출판된 논문이 현재 인용이 벌써 300건이 넘었고요.
45/365 Word Embeddings Quantify 100 Years of Gender and Ethnic Stereotypes https://t.co/CcHaWsNfU3 #365DaysofResearch #NLProc
RT @j2bryson: Word embeddings quantify 100 years of gender and ethnic stereotypes https://t.co/xYIkpvygmD Using #ai to quantify, understa…
@tokutokupooh https://t.co/4VE7zVd9BM ↑みたいなsemanticの変化を考えていて、どの程度トレンドから乖離したり、(構造変化のように)意味変化が起こるのかなどを考えていました。適当にバイアスを定義して、変化を調べる(https://t.co/z3aqHGpGwf)とかもあるけど、もっと自然に調べたいというお気持ち表明でした。
agree. also interesting comparison of adjectives. i get the feeling that 1910 was more friendly, respectful and allows different paths than did 1950s and 1990s. https://t.co/3uLAYQySIh
RT @j2bryson: Word embeddings quantify 100 years of gender and ethnic stereotypes https://t.co/xYIkpvygmD Using #ai to quantify, understa…
RT @j2bryson: Word embeddings quantify 100 years of gender and ethnic stereotypes https://t.co/xYIkpvygmD Using #ai to quantify, understa…
Word embeddings quantify 100 years of gender and ethnic stereotypes https://t.co/xYIkpvygmD Using #ai to quantify, understand & communicate who we are makes much more sense to me than trying to hack & limit who we are through digital constraints
100年間のテキストデータから、ジェンダーなどのステレオタイプに関する傾向を定量化したよという研究。 Word embeddings quantify 100 years of gender and ethnic stereotypes https://t.co/H6ChfUmfZu
RT @keisemi: #経セミ ８・９月号・海外論文サーベイ＝機械学習でテキストに隠れたバイアスを定量化する、公開！ 松井暉さん（南カリフォルニア大学CS学部博士課程）に執筆頂きました。紹介論文： https://t.co/2DyOQlKhyA 100年に及ぶ資料の文字情報…
RT @keisemi: #経セミ ８・９月号・海外論文サーベイ＝機械学習でテキストに隠れたバイアスを定量化する、公開！ 松井暉さん（南カリフォルニア大学CS学部博士課程）に執筆頂きました。紹介論文： https://t.co/2DyOQlKhyA 100年に及ぶ資料の文字情報…
Word embeddingを社会科学的な分析に使った論文のサーベイです。最後に簡単な文献紹介も入れています。
#経セミ ８・９月号・海外論文サーベイ＝機械学習でテキストに隠れたバイアスを定量化する、公開！ 松井暉さん（南カリフォルニア大学CS学部博士課程）に執筆頂きました。紹介論文： https://t.co/2DyOQlKhyA 100年に及ぶ資料の文字情報を解析しジェンダーバイアスの有無を探るとても面白い研究です！
@cnosilla This paper gets into that topic, but it's focusing on news writing vs. literature: https://t.co/p3FXzs93eO
@databoydg @brendan642 here's an example paper of my intended use of "applications of ML to show biases in society (not core ML)": https://t.co/rWJtEf9hKr
Word embeddings quantify 100 years of gender and ethnic stereotypes https://t.co/WnMF2ATxHb
There are lots of great references in that talk, but I want to highlight 2️⃣ that are really great reads: Garg et al (2018) *use* bias in word embeddings to quantify change in social attitudes 😮 https://t.co/DaDKAylmzo
@awakenting Ich hab mir den durchschnittlichen Abstand von pos. vs neg. Wörtern (Liste von Adjektiven) zu einem Bestimmen Wort z. B. Migrant oder Ausländer angesehen. Kam aber nichts richtiges raus, und hab es nicht weiter verfolgt. Ähnlich zu dem Paper: h
RT @PickersgillM: On the intersection between #machinelearning and #socialscience. https://t.co/U74bQqPwLf
On the intersection between #machinelearning and #socialscience.
'Word embeddings quantify 100 years of gender and ethnic stereotypes' https://t.co/R7xdECXEEP
This is such a creative and interesting paper. Biases in word embeddings track historical trends like labor force partipation. "We demonstrate that word embeddings can be used as a powerful tool to quantify historical trends and social change." https://
RT @gwern: @TomerUllman And people say language models don't learn anything about the real world! But word2vec's already been shown to accu…
RT @math_rachel: Excellent guest lecture by @NikhGarg on his work "Word embeddings quantify 100 years of gender and ethnic stereotypes," pu…
Word embeddings quantify 100 years of gender and ethnic stereotypes https://t.co/LfhRiDaD3N #AI #Research via @yoavgo
@dsvib @Giligg @GadiAleks @Qwigwog2 @ExMaahal @rinaarts @amirvaxman (אגב, יש את המחקר הזה, שהשתמש בגדול בשיטה שויברפון מציע, אבל לא בחן צבעים. https://t.co/rWJtEfqSBZ)
Word embeddings quantify 100 years of gender and ethnic stereotypes https://t.co/uKL3B1ad5f
RT @math_rachel: Excellent guest lecture by @NikhGarg on his work "Word embeddings quantify 100 years of gender and ethnic stereotypes," pu…
Word embeddings quantify 100 years of gender and ethnic stereotypes https://t.co/LfhRiDaD3N #AI #Research via @jeremyphoward
Nice lecture on how word embeddings quantify stereotypes https://t.co/veIvrivaBO Original Paper referenced during @NAACLHLT keynote by @random_walker https://t.co/7oDpEd0eb8
RT @math_rachel: Excellent guest lecture by @NikhGarg on his work "Word embeddings quantify 100 years of gender and ethnic stereotypes," pu…
RT @math_rachel: Excellent guest lecture by @NikhGarg on his work "Word embeddings quantify 100 years of gender and ethnic stereotypes," pu…
Word embeddings quantify 100 years of gender and ethnic stereotypes https://t.co/3Ls6sM0U8R #AI #Business via @math_rachel
Deep NLP turns around to facilitate linguistic findings. Full circle.
RT @math_rachel: Excellent guest lecture by @NikhGarg on his work "Word embeddings quantify 100 years of gender and ethnic stereotypes," pu…
RT @math_rachel: Excellent guest lecture by @NikhGarg on his work "Word embeddings quantify 100 years of gender and ethnic stereotypes," pu…
RT @NikhGarg: Thank you for having me! What a great class with a bunch of excellent questions! https://t.co/tEJbsuzkdJ
Thank you for having me! What a great class with a bunch of excellent questions!
RT @math_rachel: Excellent guest lecture by @NikhGarg on his work "Word embeddings quantify 100 years of gender and ethnic stereotypes," pu…
Excellent guest lecture by @NikhGarg on his work "Word embeddings quantify 100 years of gender and ethnic stereotypes," published in PNAS: https://t.co/1Wk7kSINpg class lecture: https://t.co/oArMWutz1E https://t.co/xSWo6UlNoS
RT @bowkerl: Hey @Evanmassenhove and @tarfandy - I stumbled across this article and thought you might find it interesting (if you don't kno…
Hey @Evanmassenhove and @tarfandy - I stumbled across this article and thought you might find it interesting (if you don't know of it already, that is!) https://t.co/k5h2faccyv #MachineBias
D. ML as a magnifying lens to study society: @NikhGarg used word vectors to study bias over 100 years for which statistical data would otherwise have not been available! https://t.co/gyDg7XeOfJ (16/)
RT @annekroon: Bias in AI is a bad thing, BUT also useful to demonstrate fluctuations in gender stereotypes over time: https://t.co/HLchXk7…
RT @annekroon: Bias in AI is a bad thing, BUT also useful to demonstrate fluctuations in gender stereotypes over time: https://t.co/HLchXk7…
Bias in AI is a bad thing, BUT also useful to demonstrate fluctuations in gender stereotypes over time: https://t.co/HLchXk7w8G #bias #AI #biasedAI https://t.co/2U6jnHoH2M
How did gender and racial stereotypes about jobs changed in US during the 20th century? https://t.co/O4xcpiuo6c https://t.co/Jm6HO1aFea
Monthly Report by Essentials: Most discussed articles about #ai: Word embeddings quantify 100 years of gender and ethnic stereotypes https://t.co/JNDmcDyA9v #machinelearning #artificialintelligence via @gwern
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
As a machine learning practitioner, it's important to watch out for biases in the data. Here's an interesting paper on 100 years of gender and race bias in words, extracted with word embeddings: https://t.co/D8kIa2nKJ7
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @bruce_2b: Using #MachineLearning to study changing stereotypes & attitudes toward women & ethnic minorities over decades: https://t.co/…
RT @bruce_2b: Using #MachineLearning to study changing stereotypes & attitudes toward women & ethnic minorities over decades: https://t.co/…
RT @bruce_2b: Using #MachineLearning to study changing stereotypes & attitudes toward women & ethnic minorities over decades: https://t.co/…
Using #MachineLearning to study changing stereotypes & attitudes toward women & ethnic minorities over decades: https://t.co/AQjYURFf7M HT @baxterkb
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @agriculios: @johnlsheridan @AnnaSexton100 @DavidUnderdown9 @safiyanoble Might be of interest: https://t.co/V3AkESfP4P @math_rachel
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
@johnlsheridan @AnnaSexton100 @DavidUnderdown9 @safiyanoble Might be of interest: https://t.co/V3AkESfP4P @math_rachel
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…
RT @math_rachel: Word embeddings quantify 100 years of gender and ethnic stereotypes paper: https://t.co/1Wk7kSINpg code: https://t.co/nBi4…