Report

📄 View Full PDF

⬇️ Download PDF


Table of Contents

  1. Introduction
  2. Background and Objective
  3. Literature Review
  4. Methodology Overview
  5. Key Findings
  6. Conclusion & Limitations
  7. References

Introduction

Background and Objective

From Snow White and Cinderella to modern classics like Frozen and Moana, Disney Princess movies have become cultural touchstones that resonate across generations. These stories are not just animated tales; they are deeply embedded in our collective memory, symbolizing everything from childhood nostalgia to evolving ideals of heroism, love, and self-discovery. As the characters and themes have evolved over the years to reflect shifting social values, so too has the public’s response.

This project seeks to delve into the rich world of Disney Princess movies by analyzing public sentiment through user-generated content on platforms like Rotten Tomatoes and IMDb. By examining how audiences engage with these iconic films—capturing sentiment, thematic discussions, and shifting perceptions over time—this study will offer a fresh look at the cultural legacy of the Disney Princess franchise. Leveraging data science tools such as sentiment analysis, clustering, natural language processing (NLP), and predictive modeling, we aim to uncover what makes these movies enduringly popular and how audience perceptions have evolved from the golden age of Disney classics to today’s empowering heroines.


Literature Review

The economic and cultural success of Disney Princess movies has been a recurring theme in academic literature. In Malfroid’s paper ‘Gender, Class, and Ethnicity in the Disney Princesses Series’, he examines the impact of Disney Princess movies during the 1990s under the leadership of Eisner after the death of Walt Disney. Eisner’s tenure, often referred to as ‘Disney Renaissance’ (1989-1999), saw several princess movies being made, including The Little Mermaid (1989), Beauty and the Beast (1991), Aladdin (1992), Pocahontas (1995), and Mulan (1998). These movies were said to ‘revive and expand the Disney brand’ by mass media; however, Malfroid argues this Disney Renaissance era, while being a huge financial success, often catered to western-centric ideologies and audiences. He critiques how non-dominant cultures were portrayed in a reductive way, with many heroines still centered on self-sacrifice and romance, rather than real self-discovery (Malfroid, 2009).

Compared to Eisner, ‘The Modern Era’ mentioned in Shahid’s 2020 paper ‘The Gender Role Portrayal of Disney Princesses and Its Impact on the Audience’ addressed the issue of ‘independent but still portraying traditional romance dynamics’ to the next level. Shahid highlights how protagonist female characters like Moana and Elsa drew audience attention with their adventurous, leadership, and self-reliant personalities, using romance as a subplot (Shahid, 2020). In both Shahid and Malfroid’s papers, the Disney Princess franchise was grouped into categories based on princess portrayal and storytelling style. For our data science project, we aim to explore the possibility of identifying these meaningful transitions and financial twists with innovative data science methods like clustering, acknowledging the importance of narrative attributes to achieve box-office success.

Since we are not using Time-Series modeling methods and are only using clustering to identify patterns between various Disney Princess movies given their corresponding features, we also aim to utilize NLP methods to identify possible emotional trends of audiences and evaluate feminism & cultural narratives through reviews. Pouransari & Ghili, in their paper “Deep Learning for Sentiment Analysis of Movie Reviews,” used 5000+ IMDB audience reviews to perform binary and multi-class sentiment classification with feature extraction methods like Bag of Words and Word2Vec. They achieved strong accuracies using Random Forest and logistic regression classifiers. Having a method backed by published research, we are confident that traditional machine learning methods (Naïve Bayes, Random Forest, KNN), being both efficient and simple, can be used for classification tasks on movie reviews. Highlighting patterns in audience preferences, such as which types of movies or narratives (e.g., feminist themes, diverse characters) correlate with higher scores, modeling classification using traditional data science techniques can help us dive further into sentiment analysis.

In the paper “The Cinderella Complex: Word embeddings reveal gender stereotypes in movies and books,” the authors use word embedding techniques to analyze cultural narratives across 6,087 movie synopses and scripts. Their findings reveal a pattern where female characters often display emotional dependency on male characters, perpetuating gender stereotypes. This underscores the relevance of analyzing reviews for hidden biases, such as those related to feminism and diversity. Similarly, in “Deep Learning for Sentiment Analysis of Movie Reviews,” deep learning models like Recursive Neural Tensor Networks (RNTN) demonstrated superior performance in capturing sentiment nuances, which further supports the integration of modern NLP techniques in audience review analysis. We believe that taking a deep dive into audience review can not only review hidden clusters in between 13 princess movies, but also reveal gender and cultural patterns of movies, and how our society’s new generation of audiences grow to be more accepting of minorities and less into stereotypes overtime. In Fu et al.’s “Measuring Cultural Representation in Films: A Computational Approach”, the team was able to combine sentiment analysis with thematic analysis to measure diversity and inclusivity in films, we intend to conduct similar studies, focus on both feminist and cultural sentiment analysis in Disney Princess movies.


Methodology Overview

Dimensionality Reduction

  • Goal: “Can we identify patterns or similarities among Disney Princess movies based on their features?” Dimensionality reduction simplifies complex data with many features by representing it in fewer dimensions while retaining its meaningful information. To ensure the accuracy of these methods, we extracted as many numerical and categorical features as possible, including cumulative box office gross, production details, and other relevant movie attributes. We used two widely accepted techniques: Principal Component Analysis (PCA) and t-distributed Stochastic Neighbor Embedding (t-SNE), which allowed us to visualize the relationships between movies in a simplified 2D space. Through our data exploration journey, we found Principal Component Analysis (PCA) to be the preferred method.

Sentiment Analysis

  • Goal: “How do audiences respond to cultural and feminist themes in Disney Princess movies?”

To analyze the representation of cultural and feminist themes in Disney Princess movie reviews, we developed an automated process to filter and assess relevant reviews by defining two sets of keywords:

  • Cultural Keywords: Words related to cultural representation, heritage, and equity (e.g., diverse, multicultural, marginalized).
  • Feminism Keywords: Words related to gender equality, female empowerment, and representation (e.g., empower, gender equality, female protagonist).

Using these keywords, we filtered reviews that mentioned cultural and feminist themes from the collected Rotten Tomatoes dataset. Looking at the average sentiment score and the number of reviews containing cultural and feminism keywords, we can identify the overall engagement and sentiment towards cultural & gender topics for each movie.

K-Means Clustering Method

  • Goal:”Can we uncover distinct themes or patterns in long reviews for Disney’s Princess Movies by grouping similar reviews into groups?”

To uncover themes in detailed reviews of Disney’s Frozen, we grouped similar reviews together based on the words used in their content. First, we filtered the dataset to focus only on longer reviews—those with at least 300 characters and a few complete sentences—so we could analyze more meaningful feedback. Then, we used a technique called TF-IDF, which helps identify the most important words in the reviews while ignoring common, less meaningful words like “the” or “is.” Next, we applied a grouping method called clustering, which automatically organizes reviews with similar content into distinct groups. To decide how many groups were best, we used a method that measures how well reviews fit into clusters. From this, we found that dividing the reviews into four groups worked well. Finally, we created word clouds for each group, which visually highlight the most frequently used words in those reviews. This helped us understand the key themes discussed by the audience.

Regression Method

  • Goal: “How well can we predict movie review scores based on the content and characteristics of reviews?” We conducted a regression task to predict numerical scores (ratings) from movie reviews by combining textual and numerical features. The primary goal was to understand how the content and characteristics of reviews influence the scores assigned by users.

Key Findings

  1. Box Office and Performance Patterns:
    Through a method of unsupervised learning, we were able to identify clusters of similar movies in terms of income through its other features. Frozen II and Frozen emerged as the highest-grossing movies, earning $1.45 billion and $1.27 billion, respectively. Other successful films like Moana and Tangled clustered together, indicating similar performance. In contrast, older movies such as Snow White and the Seven Dwarfs and Sleeping Beauty appeared farther apart, reflecting their comparatively lower box office success and older release periods.

  2. Predicting Ratings:
    We built a model to predict movie review scores based on the content and features of the reviews. Results showed that the model worked best for Beauty and the Beast, achieving high prediction accuracy. Other movies like Aladdin and Tangled also performed well, with predictions closely matching the actual ratings. Moderate accuracy was observed for movies like Cinderella and Moana, showing reliable but slightly less precise results. However, the model struggled with Pocahontas, which had lower accuracy, likely due to the greater variety in review content.

  3. Cultural Representation:
    We analyzed how audiences discussed cultural themes in their reviews. Modern Disney movies, such as Moana (2016) and Moana 2 (2024), received the most positive feedback, with cultural sentiment scores of 0.26 and 0.37, respectively. This suggests that recent films are praised for their efforts to represent cultural diversity. On the other hand, older movies like Cinderella and Snow White showed much lower engagement and positivity for cultural themes, reflecting dated portrayals of diversity. Interestingly, The Princess and the Frog stood out with the highest number of cultural-related reviews, though audience reactions were more mixed.

  4. Feminist Themes:
    Feminist themes were also explored, revealing how audiences responded to female empowerment in Disney Princess movies. Mulan (1998) led in engagement, with the highest number of feminism-related reviews, reflecting strong audience recognition of its empowering message. Beauty and the Beast (1991) stood out with the most positive feedback for feminist themes, though it had only one review mentioning this topic. Modern films like Moana and Frozen also received positive feedback, while older movies like Cinderella showed little engagement with feminist themes, aligning with their traditional portrayals of female characters.

  5. Review Length and Ratings:
    To understand how review length relates to ratings, we analyzed the data using scatter plots and boxplots. Most reviews, regardless of score, were relatively short, often under 1000 characters. However, we noticed that extreme scores—1 (very negative) and 5 (very positive)—were associated with longer reviews and greater variability. This suggests that users who feel strongly about a movie, whether positively or negatively, tend to write more detailed reviews. Scores of 2, 3, and 4, which are more neutral, had shorter and more consistent review lengths.


Conclusion & Limitations

Our project showed how Disney Princess movies have changed over time in terms of cultural representation, gender roles, and audience reception. Financially speaking, modern Disney movies like Moana and Frozen tend to earn higher than older Disney movies, with the exception of Snow White being the ground-breaking first Princess animation made by Disney, having the highest earning Disney movie, even with inflation taken into account. In terms of cultural recognition, films like Moana and Moana 2 received more positive feedback for their efforts to show diversity and heritage, matched with Shahid (2020)’s explanation that Disney’s newer movies try to connect with today’s values of cultural authenticity. On the other hand, older films like Cinderella and Snow White had much lower engagement with cultural themes. This suggests that their portrayals of culture were more limited and less relatable for modern audiences. The analysis of feminist themes tells a similar story. Movies like Mulan had the most engagement, as audiences recognized its strong message of empowerment and independence. Modern films like Moana and Frozen also performed well reflecting what Malfroid (2009) described about Disney’s effort to address gender and class struggles in recent years, and it’s failure to do so during the Disney Renaissance period. In summary, Disney Princess movies have changed to reflect cultural and social shifts. Modern movies focus more on diversity and female empowerment, and this has led to more positive responses from audiences. These changes support ideas from researchers like Shahid (2020) and Malfroid (2009), who explain how cultural representation and gender equality are becoming more important in storytelling. Disney’s ability to adapt its stories has helped it connect better with today’s audiences and create stronger, more meaningful princess characters.


References

  1. Malfroid, K. (2009). Gender, Class, and Ethnicity in the Disney Princesses Series.
  2. Shahid, M. (2020). The Gender Role Portrayal of Disney Princesses.
  3. Pouransari, H., & Ghili, H. (2021). Deep Learning for Sentiment Analysis of Movie Reviews.
  4. Fu, J., Yu, W., & Guo, J. (2021). Measuring Cultural Representation in Films.
  5. Sandler, E., et al. (2020). The Cinderella Complex.