Collecting Large-scale Comparative Text Data on Legislative Debates

Parliamentary speeches present one of the most consistently available sources of information about the political priorities, actor positions, and conflict structures in democratic states. Recent advances of automated text analysis offer more and more tools to tap into this information reservoir in a systematic manner. However, collecting the high-quality text data needed for unleashing the comparative potential of the various text analysis algorithms out there is a costly endeavor and faces various pragmatic hurdles. Against this challenge, this chapter offers three contributions. First, we outline best practice guidelines and useful tools for researchers wishing to collect or to extend existing legislative debate corpora. Second, we present an extended version of the ParlSpeech Corpus. Third, we highlight the difficulties of comparing text-as-data outputs across different parliaments, pointing to varying languages, varying traditions and conventions, and varying metadata availability.

The Pandemic and Gender Inequality in Academia

ABSTRACT Has the pandemic exacerbated gender inequality in academia? We provide real- time evidence by analyzing 1.8 million tweets from approximately 3,000 political scientists, leveraging their use of social media for career advancement. Using automated text analysis and difference-in-differences estimation, we find that although faculty members of both genders were affected by the pandemic, the shift to remote work caused women to tweet less often than their male colleagues about professional accomplishments. We argue that these effects are driven by the increased familial obligations placed on women, as demonstrated by the increase in family-related tweets and the more pronounced effects among junior academics. Our evidence demonstrating the gendered shift in professional visibility during the pandemic provides the opportunity for proactive efforts to address disparities that otherwise may take years to manifest.

EXPRESS: Emotional Expressions and Brand Status

This project investigates emotionality by brands on social media. First, we analyze a field dataset of over 200,000 text and images posts by brands across two major platforms. Using recent automated text analysis (Study 1a) and computer vision methods (Studies 1b and 1c), we provide initial documentation of a negative relationship between brand emotionality and status. Exploring this relationship further, in Studies 2, 3, and 4, we find that brands can leverage this association, reducing emotionality in brand communications to increase perceived brand status. This strategy is effective because reduced emotionality is associated with high-status communication norms, which evoke high-status reference groups. This finding is moderated by the status context of the brand (Study 2) and the product type (Study 4).

Identifying Themes and Patterns on Management of Horticultural Innovations with an Automated Text Analysis

This research provides an overview on horticulture innovations in the last decade through a literature review and the use of a computer qualitative data analysis. We used Leximancer text mining software to identify concepts, themes and pathways linked with horticulture innovations. The software tool enabled us to “zoom out” to gain a broad perspective of the pooled data, and it indicated which studies clustered around the dominant topic. It displays the extracted information in a visual form, to wit, an interactive concept map, which summaries the interconnected themes and demonstrates any interdependencies. The text mining analysis revealed that the themes strongly related to “innovation” are “water”, “urban”, “system”, “countries” and “technology”. The outputs identified have been interpreted to discover meaning from the content analysis, since the software can facilitate a comprehensive and transparent data coding but cannot replace researcher’s interpretive work. Furthermore, we focused on the diffusion and the barriers for the spread of innovation, pointing out the differences about developing and advanced countries. This analysis allows the researcher to have a holistic understanding of the examination area and could lead to further studies.

Decoupling Under Scrutiny: Consistency of Managerial Talk and Action in the Age of Nonprofit Accountability

Decoupling theory suggests inconsistencies in what nonprofits do and what they claim to do. Accountability is a potential antidote to such inconsistencies in the nonprofit sector. To test whether different features of accountability prevent decoupling, I examine the divergence in statements about managerialism among nonprofit organizations in a major U.S. metropolitan area. The analysis compares a survey of organizations to public discourse based on five-million-word website text. Professionalism and evaluation indeed prevent organizations from embellishing their discourse. However, inconsistencies between managerial practices and managerial discourse remain frequent: Organizations continue to present symbolic displays of managerialism to the general public, particularly when their missions are tangible. Furthermore, ratings generate inconsistencies by leading organizations to downplay managerial practices. This study develops an institutional understanding of managerial talk and action, shows that the problem of decoupling in the “age of accountability” is multifaceted, and has implications for the estimation of nonprofit practices using automated text analysis.

Automated Text Analysis

This chapter examines automated text analysis (ATA), which describes the different methodologies that can be applied in order to perform text analysis with the use of computer software. ATA is a computer-assisted method for analysing text, whenever the analysis would be prohibitively labour-intensive due to the volume of texts to be analysed. ATA methods have become more popular due to current interest in big data, taking into account the volume of textual content that is made easily accessible by the digitization of human activity. Key to ATA is the notion of corpus, which is a collection of texts. A necessary step before starting any analysis is to collect together the necessary documents and construct the corpora that will be used. Which texts need to be included in this step is dictated by the research question. After text collection, some processing steps need to be taken before the analysis starts, for example tokenization and part-of-speech tagging. Tokenization is the process of splitting a text into its constituent words, also called tokens, whereas part-of-speech tagging assigns each word a label that indicates the respective part-of-speech.

Shades of theory: A topic modelling of ways of theorizing in accounting history research

The article aims to explore ways of theorizing in accounting history research. The article draws on findings originating from a semi-automated text analysis by means of topic modelling of 1,300 accounting history papers published between 1996 and 2015 across six journals most relevant to the discipline. Findings show the presence of a whole range of ways of theorizing at different levels of abstraction (from narrating to conceptualizing to theorizing settings to grand theorizing). Different ways of theorizing tend to be associated not only with specific research objects but also with specific journal types. Overall, both narrating and grand theorizing are relatively decreasing in favour of mid-range theorizing approaches, which seem to be on the rise.