{"title":"Text Classification in Organizational Research – A Hybrid Approach Combining Dictionary Content Analysis and Supervised Machine Learning Techniques","authors":"Heiko Hossfeld, Martin Wolfslast","doi":"10.5771/0935-9915-2022-1-59","DOIUrl":null,"url":null,"abstract":"Big Data is an emerging field in organizational research as it provides new types of data, and technologies like digitization and web scraping allow to study huge amounts of data. Since large parts of digital data consist of unstructured text, text classification - assigning texts (or parts of texts) to predefined categories - is a central task. Text classification not only allows to identify relevant texts in a jumble of data but also to extract information from texts, such as sentiments, topics, and intentions. However, large amounts of textual data require the use of automated text mining methods, which is mostly uncharted territory in organizational research. We, therefore, outline and discuss the two existing approaches to text classification, one originating from social science (dictionary content analysis) the other from computer science (supervised machine learning). Since both approaches have advantages and disadvantages, we combine ideas from both to develop a hybrid approach that reduces existing issues and requires significantly less knowledge in programming and computer science than supervised machine learning. To illustrate our approach, we develop a classifier that identifies critical media coverage of organizational actions.","PeriodicalId":47269,"journal":{"name":"Management Revue","volume":null,"pages":null},"PeriodicalIF":7.6000,"publicationDate":"2022-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Management Revue","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.5771/0935-9915-2022-1-59","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"0","JCRName":"MANAGEMENT","Score":null,"Total":0}
引用次数: 0
Abstract
Big Data is an emerging field in organizational research as it provides new types of data, and technologies like digitization and web scraping allow to study huge amounts of data. Since large parts of digital data consist of unstructured text, text classification - assigning texts (or parts of texts) to predefined categories - is a central task. Text classification not only allows to identify relevant texts in a jumble of data but also to extract information from texts, such as sentiments, topics, and intentions. However, large amounts of textual data require the use of automated text mining methods, which is mostly uncharted territory in organizational research. We, therefore, outline and discuss the two existing approaches to text classification, one originating from social science (dictionary content analysis) the other from computer science (supervised machine learning). Since both approaches have advantages and disadvantages, we combine ideas from both to develop a hybrid approach that reduces existing issues and requires significantly less knowledge in programming and computer science than supervised machine learning. To illustrate our approach, we develop a classifier that identifies critical media coverage of organizational actions.
期刊介绍:
Management Revue - Socio-Economic Studies is an interdisciplinary European journal that undergoes peer review. It publishes qualitative and quantitative work, along with purely theoretical papers, contributing to the study of management, organization, and industrial relations. The journal welcomes contributions from various disciplines, including business and public administration, organizational behavior, economics, sociology, and psychology. Regular features include reviews of books relevant to management and organization studies.
Special issues provide a unique perspective on specific research fields. Organized by selected guest editors, each special issue includes at least two overview articles from leaders in the field, along with at least three new empirical papers and up to ten book reviews related to the topic.
The journal aims to offer in-depth insights into selected research topics, presenting potentially controversial perspectives, new theoretical insights, valuable empirical analysis, and brief reviews of key publications. Its objective is to establish Management Revue - Socio-Economic Studies as a top-quality symposium journal for the international academic community.