I ll Make a Man Out of You: A Network Analysis of Disney s Mulan Allison Gagliano, Robert Johnson, and Stefanos Stravoravdis
Background Mulan pretends to be a man and takes her father s place in the Chinese army to fight against the Huns She is accompanied by her guardian dragon, Mushu, and the rest of the Chinese army, including Shang, Ling, Yao, and Chien-Po She becomes the hero of China by defeating the Hun army and its leader, Shan-Yu
Research Questions and Hypotheses Who is the most important character(s)? Based on the betweenness centrality and the Eigenvector centrality measures, we expected Mulan to be the most important character What are the communities in Mulan? Heroes, Villains, and a couple of Neutral Character communities What is the sentiment of the following characters: Mulan, Mushu, Shang, Shan-Yu, and Yao? Mulan has mostly positive sentiment Mushu has mostly negative sentiment Shang has mostly neutral sentiment Shan-Yu has mostly negative sentiment Yao has mostly negative sentiment
Methods for Collecting Data Used R to: Extract a list of characters Extract the lines for each character by scene Create a edge list based on the character dialogue in each scene Edge list: all combinations of character interactions by scene Create a word cloud for Mulan, Mushu, Shang, Shan-Yu, and Yao Word cloud: displays frequency of unique, important words (removing stop words such as the, and, a, etc.) Perform a sentiment analysis for the 5 characters to identify: The number of positive, negative, and neutral phrases and create a bar graph from these values The most positive and most negative line for each Highlight the positive, negative, and neutral phrases, with scores, for each
Methods for Collecting Data Used Gephi to: Run Betweenness Centrality How frequently that [node] lies on short paths between other pairs of [nodes] (Beveridge, 2016) Indicates that a large amount of information passes through that character Run Eigenvector Centrality A measure of how important a node is based on its connection to other important nodes Determine communities based on modularity A community is a subgraph of similar and connected characters in the main network Modularity is a method to find optimal community divisions within a network Create Networks based on: Modularity class, with nodes colored by ranking betweenness centrality Villain/Hero comparison, with nodes colored by ranking Eigenvector centrality and positioned by degree
Challenges Script format Character labels were formatted as Mulan: or Mulan [stage direction]: Songs Inconsistency/Misspelling of names such as Long Hair Hun Man versus Hun Long Hair Guy Combining characters for labels, such as Both, All Three, Ling, Yao, and Chien-Po These issues made it difficult to create a regular expression for R We had to create new characters, separate characters, and assume characters with different labels were truly different
Results
Word Clouds and Sentiment
Mulan
Mulan Most negative: They're disgusting. Most positive: Please bring honor to us all!
Mushu
Mushu Most negative: You re worst nightmare. Most positive: All right, you might want to light that right about now. Quickly! Quickly!
Shang
Shang Most negative: I don t understand. My father should have been here. Most positive: Good luck, Father.
Yao
Yao Most negative: He thinks he s such a lady-killer! Most positive: Yes! Perfect! Now I ll pull them to safe---ty.
Shan-Yu
Shan-Yu Most negative: I tire of your arrogance old man. Bow to me! Most positive: Perfect.
Sentiment
Centrality & Modularity Data Table
Character Degree Weighted Degree Modularity Class Eigenvector Centrality Betweenness Centrality Hero/ Villain Mulan 46 129 2 1.000000 558.141697 Hero Mushu 43 105 2 0.876342 576.819078 Hero Yao 28 79 2 0.71909 56.829792 Hero Chi Fu 28 62 2 0.706712 82.108547 Hero Chorus 24 25 1 0.685255 48.21627 Neutral Fa Li 26 40 1 0.679984 71.252778 Neutral Ling 25 61 2 0.674227 45.190507 Hero Chien-Po 23 63 2 0.651081 31.163309 Hero Shang 26 70 2 0.642589 69.524634 Hero Hair Dresser 1 21 24 1 0.613175 29.990079 Neutral Hair Dresser 2 21 24 1 0.613175 29.990079 Neutral Cri-Kee 19 27 3 0.453136 90.40119 Hero Grandma Fa 17 24 1 0.436798 10.838889 Neutral Shan-Yu 20 36 2 0.396262 406.566392 Villain Army Chorus 11 11 2 0.391193 0.375 Neutral Army Men 11 11 2 0.391193 0.375 Neutral All Townspeople 14 14 1 0.386072 0 Neutral Dresser 1 14 14 1 0.386072 0 Neutral Dresser 2 14 14 1 0.386072 0 Neutral Maiden #1 14 14 1 0.386072 0 Neutral Maiden #2 14 14 1 0.386072 0 Neutral Maiden #3 14 14 1 0.386072 0 Neutral Maiden #4 14 14 1 0.386072 0 Neutral Make-up Lady 14 14 1 0.386072 0 Neutral Matchmaker 14 15 1 0.386072 0 Neutral All Soldiers 11 11 2 0.38551 0 Hero Recruit #2 11 11 2 0.38551 0 Neutral Recruit #3 11 11 2 0.38551 0 Neutral Recruits 11 11 2 0.38551 0 Neutral General Li 10 13 2 0.338919 0.410714 Hero All Recruits 9 9 2 0.322263 0 Neutral Tattoo Soldier 9 9 2 0.322263 0 Neutral Emperor 10 21 2 0.320632 1.599817 Hero Character Degree Weighted Degree Modularity Class Eigenvector Centrality Betweenness Centrality Hero/ Villain Cow 8 8 2 0.293555 0 Neutral Hun Archer 11 11 4 0.271967 69.75385 Villain Man in Crowd #1 8 8 2 0.269887 0 Neutral Man in Crowd #2 8 8 2 0.269887 0 Neutral Ancestor 3 14 16 3 0.265716 22.39643 Neutral First Ancestor 14 16 3 0.265716 22.39643 Neutral Hun Bald Man #1 7 7 2 0.241814 0 Villain Hun Bald Man #2 7 7 2 0.241814 0 Villain Fa Zhou 8 15 1 0.213816 3.809524 Neutral Ancestor 1 12 12 3 0.205912 0 Neutral Ancestor 2 12 12 3 0.205912 0 Neutral Ancestor 4 12 12 3 0.205912 0 Neutral Ancestor 5 12 12 3 0.205912 0 Neutral Ancestor 6 12 12 3 0.205912 0 Neutral Ancestor 7 12 12 3 0.205912 0 Neutral Ancestor 8 12 12 3 0.205912 0 Neutral Ancestor 9 12 12 3 0.205912 0 Neutral Fa Deng 12 12 3 0.205912 0 Neutral Townspeople 5 5 1 0.149996 0 Neutral Yi's Son 5 5 1 0.149996 0 Hero Little Brother 5 6 3 0.14652 3.85 Neutral Parade Leader 3 3 2 0.137944 0 Neutral Bath Lady 3 3 1 0.11508 0 Neutral Hun Soldier 1 1 2 0.04853 0 Villain Barry Cook 1 1 2 0.04853 0 Neutral Bald Hun Man #1 4 4 4 0.044217 0 Villain Hun Strong Man 4 4 4 0.044217 0 Villain Long Hair Hun Man 4 4 4 0.044217 0 Villain Archer Guy 4 4 0 0.030072 0 Villain Hun Long-Hair Guy 4 4 0 0.030072 0 Villain Scout #1 4 4 0 0.030072 0 Hero Scout #2 4 4 0 0.030072 0 Hero Guard 1 2 2 0.022443 0 Hero
Networks
Modularity Network 5 communities Main characters are found in pink community- regardless of heroism Remaining communities consist of close-knit characters based on their scenes Nodes colored by ranking betweenness centrality
Heroes and Villains Nodes colored by ranking Eigenvector centrality and positioned by degree Highest degree is closest to center of spiral Blue- more associations with important characters Red- fewer associations with important characters
Ideas for Future Research Analysis using characters who perform actions in scenes, rather than just those who talk Sub-communities within the communities we found Expand to other Disney movies Determine if focal character is always the most important character Other major franchises, such as the Marvel Cinematic Universe and Pirates of the Caribbean series