Discourse Analyses of Ergenekon Cases' PR Campaigns in Taraf and Zaman Newspapers

Description

This dataset accompanies a study that employed the Topic Modelling method (based on Latent Dirichlet Allocation – LDA principles) to analyze a corpus of 1,537 opinion pieces, investigating the transformation of political and security discourse in Gulenist newspapers between 2008 and 2014. The aim of the analysis was to statistically reveal the underlying themes (topics) and the clustering of key terms (topic terms) within the corpus. The simulation identified nine main topics. The results statistically confirm a significant shift in the primary security actor from the Turkish Armed Forces (TSK) to the National Intelligence Organisation (MIT). Furthermore, the study demonstrates that media campaigns supporting judicial initiatives, such as Ergenekon and Balyoz, against the TSK were orchestrated via Taraf and Zaman newspapers using a systematic "public relations" strategy. These campaigns represented not only a legal process but also an institutional and ideological struggle. The related media content analyzed in this study can be downloaded from the following link: https://drive.google.com/drive/folders/1mymTHn7SLqcQFlz3Mnb2rdutYeXgBjWA?usp=drive_link

Steps to reproduce

Data Collection: Download the corpus of 1,537 opinion pieces from the provided Google Drive link. Ensure that all articles are in a plain text or compatible format for text analysis. Preprocessing: Clean the text data: remove punctuation, numbers, and stopwords. Normalize text by lowercasing and, if necessary, stemming or lemmatization. Tokenize the text into words or phrases suitable for Topic Modelling. Topic Modelling Analysis: Apply Latent Dirichlet Allocation (LDA) using a suitable software package (e.g., Python’s gensim or R’s topicmodels). Set parameters to extract nine main topics, as identified in the study. Train the model on the preprocessed corpus and obtain topic distributions for each document. Analysis of Results: Examine the top terms for each topic to interpret the themes. Analyze the temporal evolution of topics to track shifts in discourse. Confirm patterns such as the shift from TSK to MIT as the primary security actor and the media campaigns via Taraf and Zaman newspapers. Reproducibility Notes: Ensure that all random seeds are fixed for LDA model initialization to reproduce the same topic clusters. Document the version of all software packages used. Optional: visualize topics and term distributions using word clouds, pyLDAvis, or other visualization tools.

Institutions

• Bogazici Universitesi

  Istanbul
• Recep Tayyip Erdogan Universitesi

  Rize, Rize

Categories

Licence