Installing computational social science: Facing the challenges of new information and communication technologies in social science

Abstract

Today's world allows people to connect over larger distances and in shorter intervals than ever before, widely monitored by massive online data sources. Ongoing worldwide computerization has led to completely new opportunities for social scientists to conceive human interactions and relations in unknown precision and quantities. However, the large data sets require techniques that are more likely to be found in computer and natural sciences than in the established fields of social relations. In order to facilitate the participation of social scientists in an emerging interdisciplinary research branch of "computational social science," we propose in this article the usage of the Python programming language. First, we carve out its capacity to handle "Big Data" in suitable formats. Second, we introduce programming libraries to analyze large networks and big text corpora, conduct simulations, and compare their performance to their counterparts in the R environment. Furthermore, we highlight practical tools implemented in Python for operational tasks like preparing presentations. Finally, we discuss how the process of writing code may help to exemplify theoretical concepts and could lead to empirical applications that gain a better understanding of the social processes initiated by the truly global connections of the Internet era.

A paradigm shift in German family policy: Applying a topic model to map reform and public discourse, 1990–2016

“…I used topic modeling to identify thematically coherent topics in the large text corpus (Blei, ). Topic models are part of a family of methods that have been labeled computational social sciences (CSS) (Heiberger & Riebling, ). CSS provides a toolkit of methods aimed at processing large and often relatively unstructured data, as is the case with newspaper archives (Heiberger & Riebling, ).…”

“…Topic models are part of a family of methods that have been labeled computational social sciences (CSS) (Heiberger & Riebling, ). CSS provides a toolkit of methods aimed at processing large and often relatively unstructured data, as is the case with newspaper archives (Heiberger & Riebling, ). The underlying algorithm known as latent Dirichlet allocation (LDA) (Blei, Ng, & Jordan, ), data preparation, and my model evaluation strategy are described in Appendix .…”

Sandkastengespräche im Netz? Leitbilder „guter Erziehung“ in einem digitalen Elternforum

“…Das methodische Vorgehen kombiniert zwei Verfahren. Zunächst bediene ich mich mit "topic models" eines Ansatzes der quantitativen Textanalyse, welcher dem Repertoire der "computational social sciences" (CSS) entstammt (Heiberger/Riebling 2016). Diese Verfahren erlauben es, große und oftmals unstrukturierte Datenmengen zu verarbeiten (ibid.).…”

Sentiment analysis through twitter as a mechanism for assessing university satisfaction

“…As indicated, data mining encompasses a series of modeling techniques oriented to different requirements of organizations, with respect to the educational sector, the use of these techniques can be seen in a series of works aimed at improving the quality of university education [11], [12]. Within data mining, there is text mining, which is the process of finding implicit patterns, previously unknown, that can be useful from a text repository [13]. Likewise, text mining consists of the search for regularities or patterns found in a text, from machine learning techniques, therefore, it is considered as one of the many branches of computational linguistics [14].…”