“Krymnash” on YouTube

Abstract

The dataset was collected in three phases using snowball-sampling during 2020. The experiment was designed to simulate the situation in which users want to inform themselves about Crimea’s annexation on YouTube, and progress by watching videos relevant to the question, picking only relevant recommendations from the list of recommendations in the three stages of the experiment.

For the first data collection, fifteen starting videos were manually selected by the author using the YouTube search function. They had to be relatively short (under 20 minutes), seven of them represented a clearly positive attitude towards the annexation, seven evaluated this event negatively, and one video represented an ambivalent perspective. Collected recommendations were manually annotated. Two independent annotators watched all recommended videos in full length and evaluated such parameters as relevance, evaluation of the annexation, language, topic, and category. In addition, the presence of actual topics, such as the increasing issue of the pandemic and various anniversaries, among them the anniversary of the Crimea annexation, was marked in the dataset.

The second data collection starts from a selection of videos taken from the first and uses the foremost relevant recommendations, each recommended more than 12 times.

For the third data collection, the recommendations were collected after watching the top of relevant recommendations (44 videos, including several watched videos from the second data collection as they were frequently recommended again).

The dataset includes: 1. first dataset of collected recommendations with annotation (DATASET_1.xlsx), 2. second dataset of collected recommendations (DATASET_2.xlsx), 3. third dataset of collected recommendations (DATASET_3.xlsx), 4. Jupyter Notebook with data exploration and visualisations as JN (Krymnash_notebook_EH.jpynb, executable) and PDF (Krymnash_notebook_EH.pdf, non-executable, only readable), 5. four Gephi-graphs as visualisations of first (graph_first_dataset.gephi), second (graph_second_dataset.gephi), third (graph_third_dataset.gephi) and consolidated (1-3) data sets (graph_mutual_relations_across_datasets.gephi), 5. table with the description of stages and user variables (overview_stages_and_users.xlsx), 6. codebook for annotation of the first dataset (Codebook_annotation_dataset_1.pdf), 7. screenshots of graphs, used as visualisations in the Jupyter Notebook (category.png, engagement.png, evaluation.png, evaluation_second_ds.png, mutual_relations_network.png, relevance_third_ds.png, selection_second_ds.png) 8. properties of unique videos of all datasets (all_datasets_df_with_tags.xlsx).