Text mining for social science – The state and the future of computational text analysis in sociology

Macanovic, Ana

doi:10.1016/j.ssresearch.2022.102784

Cited by 41 publications

(19 citation statements)

References 115 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The RMD was created from the Macquarie Thesaurus (Mohammad and Turney 2013) which includes a rich set of general phrases pertaining to emotions, but lacks words specific to the emotion expression in the informal online context our short texts come from. These results echo the notes of caution that RMDs should only be used when researchers are certain of the similarity of their data to the data the RMD was created and validated on (Macanovic 2022;Wu et al 2018).…”

Section: Step 2: Methods Choicesupporting

confidence: 70%

“…However, when working with previously unmapped concepts or data from a new domain (or type of text), researchers need to create a custom-made dictionary (CMD) from the ground up (Bonikowski and Gidron 2016;Macanovic 2022;Nelson et al 2021;Spörlein and Schlueter 2021). We suggest a simple approach that uses manually coded data to automatically create such dictionaries relying on the concept of word keyness in computational linguistics .…”

Section: Step 2-methods Choicementioning

confidence: 99%

“…As such, they tend to lack variables of sociological interest or information about the context in which the texts were generated (Boyd and Schwartz 2021;Macanovic 2022). Some of these challenges can be overcome by combining textual data with other data sourcessuch as survey, census, or experimental data (Boyd and Schwartz 2021;van Loon et al 2020;Salganik 2017).…”

Section: Conclusion and Final Remarksmentioning

confidence: 99%

“…Sociological research has, so far, mostly relied on using pre-existing dictionary resources and unsupervised machine learning methods to seek concepts of interest in text (Macanovic 2022) and focused less on methodological decisions and choices (although, see Nelson et al 2021).…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

A Systematic Evaluation of Text Mining Methods for Short Texts: Mapping Individuals’ Internal States from Online Posts

Macanovic¹,

Przepiorka²

2022

Preprint

View full text Add to dashboard Cite

Sociologists have successfully used text mining to investigate discourse using news articles, official documents, and other sources. Yet, the potential of exploring millions of short texts generated spontaneously by individuals in online environments has remained untapped within the field. To fill this gap, we show how such texts can inform sociologists about individual internal states such as norms, motives, and stances, which thus far have been mainly elicited using surveys. We assess the performance of 581 variations of three text mining approaches–dictionary methods, supervised, and unsupervised machine learning–against the benchmark of texts coded by humans for complex schemes capturing individuals’ internal states. Our analysis includes coding feedback texts from an online market for motives for leaving feedback (N = 2,000) and tweet texts for moral values expressed in text (N = 3,832). We describe challenges arising with these different approaches and provide best-practice advice for future applications.

show abstract

Section: Step 2: Methods Choicesupporting

confidence: 70%

Section: Step 2-methods Choicementioning

confidence: 99%

Section: Conclusion and Final Remarksmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

A Systematic Evaluation of Text Mining Methods for Short Texts: Mapping Individuals’ Internal States from Online Posts

Macanovic¹,

Przepiorka²

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Como o objetivo geral dessa investigação é a análise das redes de interação dos usuários mais expostos à mis/desinformação/controvérsia sobre vacinas no âmbito da pandemia COVID-19, faz-se necessário identificá-la na conversação capturada. Contudo, os dados coletados perfazem quase 24 milhões de tweets, e em razão da inviabilidade de adotar qualquer procedimento manual, fez-se uso de técnicas da Mineração de Textos; uma sub-área da Inteligência Artificial que objetiva extrair informações e padrões em conjuntos de dados textuais (Tao;Hotho;et al, 2005), a qual tem sido empregada, tal qual explicitado por Jung e Lee (2020), em áreas como Educação (Ferreira-Mello et al, 2019), Saúde (Tao;, Ciências Biológicas (Cohen & Hunter, 2008) e Ciências Sociais (Macanovic, 2022).…”

Section: Metodologiaunclassified

Câmaras de eco e mis/desinformação sobre vacinas contra a COVID-19 no Twitter: abordagem baseada em machine learning e análise de redes sociais

Lopes

Brotas²

2023

RSD

View full text Add to dashboard Cite

Haja visto a importância infodemiológica das câmaras de eco na disseminação de mis/desinformação, objetivou-se analisar as redes de interação dos usuários mais expostos à mis/desinformação ou controvérsia sobre vacinas no âmbito da pandemia COVID-19. Para tal, é proposto nesta investigação uma metodologia baseada em machine learning e Análise de Redes Sociais para detecção automatizada de conteúdo controverso e mis/desinformativo sobre vacinas, através da qual chegou-se a um modelo com 92% de acurácia. Dos quase 24 milhões de tweets coletados, 12.4 milhões (52%) foram assinalados como controversos e/ou potenciais à mis/desinformação, sendo os meses de janeiro e junho de 2021 aqueles de maior atividade, sendo analisados através de uma coorte. Diferentemente dos trabalhos anteriores, analisou-se a rede de todas as maneiras de interagir no Twitter, e de toda a estrutura textual dos tweets – não apenas links ou hashtags -. No que concerne à conversação sobre vacinas contra COVID-19, os achados foram diferentes daqueles associados à discussão político-partidária anteriormente descritos na literatura, uma vez que a rede de menções e respostas privilegia relações heterofílicas, e que conformações de “eco” não foram observáveis. Finalmente, faz-se necessário novos estudos para melhor compreender a disseminação de desinformação acerca das vacinas no Twitter.

show abstract

Shifts in national entrepreneurial culture: The promise of linguistic cultural artifacts and machine learning analysis

Kyprianou,

Vedula,

et al. 2024

Strategic Entrepreneurship

View full text Add to dashboard Cite

Research SummaryWe develop a dynamic view of national entrepreneurial culture by examining the linguistic evolution of media‐produced cultural artifacts—entrepreneurship‐related newspaper articles. Applying machine learning to 690,088 articles from 103 newspapers across the United States between 1996 and 2016, we identify a growing positivity bias toward entrepreneurship at the national level evidenced by rising emotional tone and declining analytical thinking. This bias varies by topic, with “entrepreneurial aspirations and journeys” driving the trend. Our analyses also suggest this bias may encourage the creation of new ventures but limit venture growth potential. We highlight theoretical and methodological contributions to research on national entrepreneurial culture and identify promising avenues for future research.Managerial SummaryWe examine how a country's cultural attitudes toward entrepreneurship change over time by studying relevant newspaper articles. We also consider if any changes in such attitudes may have implications for the quantity and quality of a country's new ventures. After analyzing 690,088 articles from 103 newspapers across the United States between 1996 and 2016, we find a growing positivity bias toward entrepreneurship evidenced by increasing rates of positive tone and decreasing rates of analytical thinking. This bias is largest when media articles discuss entrepreneurial aspirations and journeys. Our analyses also suggest this bias may facilitate the creation of new ventures but limit their growth potential. These findings have implications for understanding and measuring national entrepreneurial culture, and create opportunities for future research.

show abstract

Text mining for social science – The state and the future of computational text analysis in sociology

Cited by 41 publications

References 115 publications

A Systematic Evaluation of Text Mining Methods for Short Texts: Mapping Individuals’ Internal States from Online Posts

A Systematic Evaluation of Text Mining Methods for Short Texts: Mapping Individuals’ Internal States from Online Posts

Câmaras de eco e mis/desinformação sobre vacinas contra a COVID-19 no Twitter: abordagem baseada em machine learning e análise de redes sociais

Shifts in national entrepreneurial culture: The promise of linguistic cultural artifacts and machine learning analysis

Contact Info

Product

Resources

About