I recently submitted my chapter in ICA Handbook of Computational Communication Research to the editors recently. Among the wide range of topics covered in the volume, my chapter, Time-dynamic Analysis, explains how to analyze textual data collected from over an extended period:
In communication research, scholars often analyze news articles, speech transcripts or social media posts to reveal how important public issues are discussed. They employ computational tools to examine how media content changes over an extended time period, but this remains challenging because the relationship between words and concepts shifts across different time periods. Therefore, understanding historical media discourses requires methods that account for temporal changes in word meanings and usages. To achieve this, scholars should employ time-dynamic analysis, in which machine learning models are applied to media content from specific time windows. This chapter explains how to design research using longitudinal media data and how to apply computational tools to perform dynamic analysis. It argues that cross-sectional time-varying analysis is particularly well-suited to computational communication research and demonstrates how dynamic analysis can be implemented using a range of unsupervised or semi-supervised machine learning models for document scaling and classification.
I am thankful to Mario Haim and Emese Domahidi for editing the volume and honored to be one of the contributors of the volume along with my great international colleagues.
