Suchergebnisse

AbstractPolitical scientists increasingly use supervised machine learning to code multiple relevant labels from a single set of texts. The current "best practice" of individually applying supervised machine learning to each label ignores information on inter-label association(s), and is likely to under-perform as a result. We introduce multi-label prediction as a solution to this problem. After reviewing the multi-label prediction framework, we apply it to code multiple features of (i) access to information requests made to the Mexican government and (ii) country-year human rights reports. We find that multi-label prediction outperforms standard supervised learning approaches, even in instances where the correlations among one's multiple labels are low.

Political scientists increasingly use supervised machine learning to code multiple relevant labels from a single set of texts. The current "best practice"of individually applying supervised machine learning to each label ignores information on inter-label association(s), and is likely to under-perform as a result. We introduce multi-label prediction as a solution to this problem. After reviewing the multi-label prediction framework, we apply it to code multiple features of (i) access to information requests made to the Mexican government and (ii) country-year human rights reports. We find that multi-label prediction outperforms standard supervised learning approaches, even in instances where the correlations among one's multiple labels are low.

How many agendas has Xi Jinping put forth and promoted since taking office in 2012, and what are the types of agendas? How much of political attention has each agenda received, and how has the allocation of attention changed over time? We utilize a dataset of presidential statements, speeches, and reports from 2012 to 2022 and employed automated text analysis to identify major topics and terms associated with each topic. Our analysis reveals the identification of about 25 distinct policy agendas across diverse policy domains, with remarkable temporal variations between agendas in terms of the amount of leadership attention. Particularly, we find a significant shift in both the substance and relative weight of policy agendas between the first and second terms of Xi's tenure, indicating his adaptation and responses to changing domestic and foreign policy environments.

The article is about the criteria used to decide the relevance of particular contexts to the framework for qualitative text analysis. By conceptualizing context as something that needs to be justified, rather than identified, we argue in favour of transparency when it comes to the link between text-as-data and interpretations of text-as-data. By drawing on examples from a speech made to the European Parliament, we examine a number of practical tools for anchoring context to text, these spanning from the literal, through cues, to absence. By means of the ideal of transparency we hope to move the focus of the critique of analysis from correct/incorrect methodology to credible/untenable interpretations.

AbstractParticipants that complete online surveys and experiments may be inattentive, which can hinder researchers' abilityto draw substantive or causal inferences. As such, many practitioners include multiple factualor instructional closed-ended manipulation checks to identify low-attention respondents. However, closed-ended manipulation checks are either correct or incorrect, which allows participants to more easily guess and it reduces the potential variation in attention between respondents. In response to these shortcomings, I develop an automatic and standardized methodology to measure attention that relies on the text that respondents provide in an open-ended manipulation check. There are multiple benefits to this approach. First, it provides a continuous measure of attention, which allows for greater variation between respondents. Second, it reduces the reliance on subjective, paid humans to analyze open-ended responses. Last, I outline how to diagnose the impact of inattentive workers on the overall results, including how to assess the average treatment effect of those respondents that likely received the treatment. I provide easy-to-use software in R to implement these suggestions for open-ended manipulation checks.

AbstractThis study analyzes supplier codes of conduct of multinational firms, with two main research objectives: (1) providing a description of supplier codes' content provisions, specifically focusing on the labor standards provisions included in these self‐regulatory policies, and (2) comparing code content across regions and sectors. To this end, the study uses hand‐coding and novel text‐as‐data techniques for content analysis of a large sample of 880 codes of conduct. Results show that a standardized list of labor rights is included in up to 90% of the aforementioned codes, regardless of the location or sector of the drafting company. Codes are drafted with a clear influence from internationally recognized standards, even though a minority of codes directly refer to international texts. However, the similarity of codes is limited as they differ in length and extent to which they elaborate certain topics. This latter aspect is correlated with the firms' location and the sector they operate in. The research demonstrates that European companies refer to the legal framework and international standards extensively, while American companies more often develop their corporate ethical values or focus on governance and their relationship with suppliers. It also empirically shows that companies evolving in reputation‐sensitive sectors are developing more specific codes including more detailed labor provisions.

Every year at the United Nations, member states deliver statements during the General Debate discussing major issues in world politics. These speeches provide invaluable information on governments' perspectives and preferences on a wide range of issues, but have largely been overlooked in the study of international politics. This paper introduces a new dataset consisting of over 7,300 country statements from 1970–2014. We demonstrate how the UN General Debate Corpus (UNGDC) can be used to derive country positions on different policy dimensions using text analytic methods. The paper provides applications of these estimates, demonstrating the contribution the UNGDC can make to the study of international politics.

The resurgence of deliberative institutions in the developing world has prompted a renewed interest in the dynamics of citizen engagement. Using text-as-data methods on an original corpus of village assembly transcripts from rural Tamil Nadu, India, this paper opens the "black box" of deliberation to examine the gendered and status-based patterns of influence. Drawing on normative theories of deliberation, this analysis identifies a set of clear empirical standards for "good" deliberation, based on an individual's ability both to speak and be heard, and uses natural language processing methods to generate these measures. The study first shows that these assemblies are not mere "talking shop" for state officials to bluster and read banal announcements, but rather, provide opportunities for citizens to challenge their elected officials, demand transparency, and provide information about authentic local development needs. Second, the study finds that across multiple measures of deliberative influence, women are at a disadvantage relative to men; women are less likely to speak, set the agenda, and receive a relevant response from state officials. Finally, the paper shows that although quotas for women on village councils have little impact on the likelihood that they speak, they do improve the likelihood that female citizens are heard.

"From social media posts and text messages to digital government documents and archives, researchers are bombarded with a deluge of text reflecting the social world. This textual data gives unprecedented insights into fundamental questions in the social sciences, humanities, and industry. Meanwhile new machine learning tools are rapidly transforming the way science and business are conducted. Text as Data shows how to combine new sources of data, machine learning tools, and social science research design to develop and evaluate new insights.Text as Data is organized around the core tasks in research projects using text--representation, discovery, measurement, prediction, and causal inference. The authors offer a sequential, iterative, and inductive approach to research design. Each research task is presented complete with real-world applications, example methods, and a distinct style of task-focused research. Bridging many divides--computer science and social science, the qualitative and the quantitative, and industry and academia--Text as Data is an ideal resource for anyone wanting to analyze large collections of text in an era when data is abundant and computation is cheap, but the enduring challenges of social science remain." --Page 4 of cover