Topicalizer: reframing core concepts in machine learning visualization by co-designing for interpretivist scholarship
Topicalizer: reframing core concepts in machine learning visualization by co-designing for interpretivist scholarship
复制标题
Topicalizer:通过解释主义学术的共同设计重新构建机器学习可视化的核心概念
DOI:
10.1080/07370024.2020.1734460
复制
发表时间:
2020
期刊:
影响因子:
--
通讯作者:
McGee, Micki
中科院分区:
文献类型:
--
作者:
Baumer, Eric P.;Siedel, Drew;McDonnell, Lena;Zhong, Jiayun;Sittikul, Patricia;McGee, Micki
Designing computational systems means, increasingly, designing for interactions with algorithms. Machine learning, recommender systems, sentiment analysis, and other algorithmic techniques have become common in interactive, user-facing systems. These developments raise novel, complex challenges, both in the design of such systems (Baumer, 2017; Dove et al., 2017; Leahu, 2016; Yang et al., 2018) and in terms of users’ interactions with them (Ananny, 2011; Bucher, 2017; Eslami et al., 2016, 2015; Gillespie, 2012, 2013). Computational tools for supporting interpretivist text analysis by sociological and humanistic scholars provides a compelling context in which to explore these issues. Given a large corpus, how does one determine what the documents are about? Traditionally, methods to address this question include qualitative analysis (Lofland et al., 1971 (2005)), grounded theory (Charmaz, 2006; Glaser & Strauss, 1967), discourse analysis (Foucault, 1972), and others. However, it can be difficult to analyze copious volumes of unstructured text via such “close reading” methods, which are time-and laborintensive.One increasingly common approach leverages computational techniques in areas such as digital humanities (Jockers, 2013; Jockers & Mimno, 2013; Rhody, 2013; Underwood, 2019, 2014) or computational social science (Lazer et al., 2009; Roberts et al., 2014). Researchers employ sentiment analysis (Pang & Lee, 2008), topic modeling (Blei, 2012; Blei et al., 2003), collocate analysis (Lind & Salo, 2002), and a variety of other techniques to analyze large volumes of text. Furthermore, computational methods do more than simply scale up human ability. Humans and computers read texts in different ways (Burrell, 2016; Passi & Jackson, 2017; Taylor, 2009). Computational systems provide an alternative lens, a different view of textual data that might not have been achieved by human researcher (s) alone (Baumer et al., 2017; Muller et al., 2016). Put differently, computational analysis is not simply faster but rather a fundamentally different means of text analysis.
登录
查看更多内容
DOI:
10.7551/mitpress/9780262525374.003.0009
发表时间:
2013
期刊:
Theory, Culture & Society
影响因子:
--
作者:
Tarleton Gillespie;P. Boczkowski
通讯作者:
P. Boczkowski
DOI:
10.2307/1423857
发表时间:
1971
期刊:
--
影响因子:
--
作者:
Toby S. Levy;J. Lofland
通讯作者:
Toby S. Levy;J. Lofland
DOI:
10.1145/2598510.2600879
发表时间:
2014
期刊:
Proceedings of the 2014 conference on Designing interactive systems
影响因子:
--
作者:
L. Leahu;Phoebe Sengers
通讯作者:
Phoebe Sengers
影响因子:
7.9
作者:
R. Lind;Colleen Salo
通讯作者:
Colleen Salo
DOI:
10.22230/src.2013v4n3a121
发表时间:
2013
期刊:
Scholarly and Research Communication
影响因子:
--
作者:
Lauren F. Klein;Jacob Eisenstein
通讯作者:
Jacob Eisenstein