Digital Humanities · YouTube Comments · Space Narratives
A comparative sentiment analysis of audience responses to animal and human space mission tragedies using TextBlob and Twitter-RoBERTa.
Research Question
The study compares two kinds of public memory around space tragedy: one centered on Laika, the first dog sent into space, and one centered on Vladimir Komarov, the Soviet cosmonaut who died during a space mission. The dashboard keeps both the narrative difference and the model difference visible at the same time.
TextBlob uses polarity scores to classify comments as positive, neutral, or negative. It gives a broad overview but can miss context, indirect grief, sarcasm, or subtle emotional language.
Twitter-RoBERTa is a contextual model fine-tuned for social media language. It captures more negative emotional tone in both datasets, especially in the Laika comments.
Interactive Explorer
Use the controls to view the sentiment distribution for each analysis run. Counts and percentages come from the executed Colab notebook outputs.
Comparative Dashboard
The comparison views show how the same YouTube comments shift when interpreted through a rule-based model versus a context-aware transformer model.
| Narrative | Model | Type | Sentiment | Count | Percentage |
|---|
Interpretation
The transformer model identifies the Laika dataset as overwhelmingly negative, with 71.50% of comments classified as negative. This suggests that animal suffering and ethical memory around space experimentation produces especially intense emotional response.
In the human dataset, TextBlob places 53.51% of comments in the neutral category. This indicates that many comments may be descriptive, historical, informational, or reflective rather than directly emotional.
TextBlob tends to give a more balanced or neutral distribution. Twitter-RoBERTa consistently increases the negative share, showing that contextual models can reveal emotional intensity that polarity-based scoring may underrepresent.
The same comment culture can look different depending on the computational lens. This is important for Digital Humanities, because methods do not simply measure emotion; they shape how emotional memory becomes visible.
Methodology
YouTube comment datasets for Laika and Vladimir Komarov were loaded in Google Colab using Pandas.
Emojis, non-ASCII characters, numbers, and repeated spaces were removed to standardize the text.
Only English-language comments were retained using language detection so that both models received comparable input.
TextBlob classified comments using polarity thresholds: positive above 0.1, negative below -0.1, and neutral in between.
Twitter-RoBERTa processed comments in batches with a short maximum token length suited to YouTube comments.
Plotly-based interactive charts were used to compare sentiment distributions across narratives and models.
Limitations
The Laika dataset is much larger than the Komarov dataset. This may make the Laika results more stable and the Komarov results more sensitive to variation.
Only English comments were analyzed. This excludes multilingual audience responses and may not represent the full global comment culture around these videos.