This is an executive summary of a text mining analysis project I did for my Audience Insight Analysis class!
Disclaimer: All content found on this page was made for academic purposes only. I have no association with any of the companies listed here.
Research Question:
How are people discussing the Hannah Montana 20th Anniversary Special online?
What are the emotions and tones driving digital conversations?
Key Conclusion:
The majority of online conversation surrounding Hannah Montana’s 20th Anniversary special centered around nostalgia for Hannah Montana, as well as for mid-2000s Disney Channel shows more broadly.
Methods:
Data Collection
Using Mention, we collected data from social media sites, such as Instagram and X/Twitter, as well as news sites and blogs.
- Search Window: Apr. 6 – Apr. 7, 2026
- n = 4,871
Pre-processing
We imported our data into Orange.
We made a list of stop-words that filtered out words and phrases irrelevant to our research question, such as online usernames. We also filtered out words like “HM,” which referred to the topic itself instead of the nature of conversations.
A large portion of our dataset included retweets, which affected our analysis. However, we believe that the nature of the retweets is still reflective of online sentiment regarding the HM20 Special, which we will discuss in subsequent sections.
Analytical Methods
Word Cloud
A visual model for text analysis, with word size corresponding to presence in the dataset.
Sentiment Analysis
Classifies data into six basic emotions: anger, disgust, fear, joy, surprise, and sadness.
Topic Modeling
“Applies unsupervised learning on large sets of texts to produce a summary set of terms that represent the collection’s overall primary set of topics.” (IBM)
Semantic Network Analysis
“A method that maps relationships between words or concepts as a network (graph), rather than just counting how often words appear.” (Claude)
Results:
Word Cloud
Our initial word cloud of the entire data set is a bit too broad to get any specific interpretations. We’ll need to do some further analysis to find more meaning.
Sentiment Analysis
- Most negative sentiment (anger, disgust, fear, sadness) were either:
- Misclassified by Orange and weren’t actually negative.
- Tweet replies without any context as to prior conversation.
- Positive sentiment came from a large amount of ‘Mount Rushmore’ retweets
While we struggled to draw any meaningful conclusions from negative emotions, the positive emotions strongly indicated a tone of nostalgia that defined online conversations
Topic Modeling
Out of four generated topics, we identified the two most insightful ones:
Topic 1: Other Disney Stars, like Selena Gomez & Raven Symone
Topic 2: Old Disney actors & hype around the March release
Semantic Network Analysis
The strongest connections were formed between terms associated with Disney Channel stars and Hannah Montana-related nostalgia, indicating that audience engagement was largely fueled by shared memories of Disney Channel culture. ?
The word “special” appeared more isolated and connected primarily through second-degree relationships, suggesting that audiences focused more on the celebrities and nostalgic associations surrounding the event rather than the event itself.
Recommendations:
1. Make Use of Nostalgic Content
- Develop social media campaigns based on themes like “younger me,” childhood memories, fan identity, etc.
- Promote iconic scenes, lyrics, and characters from all the shows
- Make users generate their own content (e.g., “Where were you when…?)
1a. Change Your Message from Special to Nostalgic Experience
- Pay less attention to the “special” itself and more to what it means for fans
- Explore themes of development, identity, and generational experience
- Tell stories based on fans’ personal experiences
2. Capitalize on the Whole Disney Channel Universe
- Use other popular Disney Channel stars in marketing strategies
- Stress connections between different shows (e.g., Raven, Selena, Dylan, and Cole)
- Develop content based on the entire Disney Channel ecosystem
Recommendations For Future Research
- Explore machine learning techniques to:
- Account for contextual features (must be complemented with more robust data collection)
- Alleviate misclassification errors