定种族:在词嵌入中改进种族维度的构建
Nnaemeka Ohamadike1, Kevin Durrheim2, Mpho Primus3
1Centre for Applied Data Science, University of Johannesburg, Johannesburg, South Africa.
概括
这项研究引入了统计指标来评估词语嵌入中的种族维度,发现名称有效地构建了这些维度. 撒哈拉以南非洲/欧洲名称嵌入显示了最稳定和最可概括的结果来衡量种族偏见.
科学领域:
- 计算语言学计算语言学
- 自然语言处理 (NLP) 是一种自然语言处理.
- 社会偏见检测检测社会偏见检测
背景情况:
- 词嵌入对于检测语言中的社会偏见是有价值的.
- 在嵌入式中测量种族偏见的研究落后于性别偏见研究,原因是复杂的种族维度构建.
- 现有的种族偏见方法依赖于主观词选择.
研究的目的:
- 开发统计指标来评估词语嵌入中的种族类别的质量和适应性.
- 解决在构建和评估种族维度方面的方法差距.
- 为偏差测量提供一个强大的种族维度构造的框架.
主要方法:
- 应用统计指标来评估三个不同的词嵌入 (谷歌新闻,南非新闻,维基百科) 的种族类别.
- 使用名称类别 (例如,撒哈拉以南的非洲人/欧洲人,美国人) 来构建种族尺寸.
- 通过将嵌入维度与人类评级相关联来验证指标.
主要成果:
- 名称类别在构建种族尺寸方面被证明是有效的.
- 与其他类别相比,撒哈拉以南的非洲/欧洲名称类别产生了更稳定和更可概括的维度.
- 美国名称类别表现出较少的概括性.
- SSA/欧洲名称嵌入显示了与人类评级的最强相关性,验证了指标捕捉感知语义结构的能力.
结论:
- 开发的统计指标为在词嵌入中构建种族维度提供了强大的框架.
- 这一框架提高了在NLP模型中准确测量种族偏见的能力.
- 这些发现强调了特定名称类别的有效性,以可靠地评估种族偏见.
相关概念视频
RACE - Rapid Amplification of cDNA Ends
7.2K
Rapid Amplification of cDNA Ends, or RACE, is one of the most effective methods to obtain a full-length cDNA from an mRNA sequence between a known internal region to the unknown sequence at the 5’ or 3’ end. The unknown region is cloned in the cDNA by a gene-specific primer that binds the known end, and a hybrid primer that attaches a predefined anchor sequence to the unknown end of the cDNA. The sequence in between is amplified by PCR with an anchor primer and a gene-specific...
7.2K
11:48A Contrast of Three Inoculation Techniques used to Determine the Race of Unknown Fusarium oxysporum f.sp. niveum Isolates
3.7K
Managing Fusarium wilt of watermelon requires knowledge of the pathogen races present. Here, we describe the root-dip, infested kernel seeding, and modified tray-dip inoculation methods to demonstrate their efficacy in race-typing of the pathogenic fungus Fusarium oxysporum f. sp niveum...
3.7K
08:13Using the Race Model Inequality to Quantify Behavioral Multisensory Integration Effects
6.8K
The current study aims to provide a step-by-step tutorial for calculating the magnitude of multisensory integration effects in an effort to facilitate the production of translational research studies across diverse clinical...
6.8K
Design Example: Dimensioning of Concrete Masonry Construction
289
For the construction of a storeroom using concrete masonry units, it's essential to align the dimensions of the structure with the actual sizes of the blocks and the intended mortar joints. On the site in question, there's a stockpile of concrete masonry blocks with a nominal size of eight by eight by sixteen inches, which are to be used in the construction of the storeroom.
The site engineer has laid out a plan for the storeroom with external dimensions of twelve feet in length and...
The site engineer has laid out a plan for the storeroom with external dimensions of twelve feet in length and...
289
The Anchoring-and-Adjustment Heuristic
7.7K
In order to make good decisions, we use our knowledge and our reasoning. Often, this knowledge and reasoning is sound and solid. However, sometimes, we are swayed by biases or by others manipulating a situation. For example, let’s say you and three friends wanted to rent a house and had a combined target budget of $1,600. The realtor shows you only very run-down houses for $1,600 and then shows you a very nice house for $2,000. Might you ask each person to pay more in rent to get the...
7.7K
Lipids as Anchors
7.2K
In the plasma membrane, the lipids forming the bilayer can also act as an anchor to tether proteins to the membrane. The three main types of lipid anchors found in eukaryotes are – prenyl groups, fatty acyl groups, and glycosylphosphatidylinositol or GPI groups. Prenyl and fatty acyl groups act as anchors on the cytosolic surface of the membrane, whereas GPI anchors proteins on the extracellular side.
The carboxy-terminal of most of the prenylated proteins, such as Ras proteins, contains...
The carboxy-terminal of most of the prenylated proteins, such as Ras proteins, contains...
7.2K

