Introduction
Emotion in media is rarely produced by a single element. In both music and film, emotional atmosphere emerges from a combination of sensory cues such as tone, pacing, and descriptive framing. Studio Ghibli films are often associated with rich emotional worlds because of the way their stories and imagery establish mood through narrative language rather than relying on a single emotional signal.
This project focuses on how emotional atmosphere is constructed across two different mediums by comparing Spotify audio features and sentiment patterns found in Studio Ghibli film descriptions. Both forms create emotional tone through sensory cues, one through sound and the other through narrative language. By examining these datasets, the goal is not to assume that music and film express emotion in fundamentally different ways, but to explore what patterns emerge when comparing their emotional signatures are analyzed together.
To guide this descriptive analysis, I focus on three central questions:
1. What do the emotional and sensory profiles of most streamed Spotify tracks in 2023 look like?
How are features like valence(brightness), energy, and acousticness distributed across these songs?
2. How do these musical features relate to one another?
Are bright songs also high-energy? Do warmer, acoustic songs tend to have lower loudness or tempo?
3. How do these musical emotional patterns compare to the emotional themes found in Studio Ghibli film descriptions?
Do songs convey emotion more intensely, while Ghibli descriptions show more balance and nuance?
Primary Source
The primary dataset for this project is a collection of Spotify audio features for most streamed tracks of 2023. Each row represents a single song and includes variables that describe the musical and emotional qualities of the track, such as valence, energy, acousticness, tempo, and danceability. These features allow each song to be described in terms of how it sounds and feels, rather than by popularity metrics alone.
This dataset is especially well suited for emotional analysis because Spotify’s audio features compress complex musical signals into interpretable values. Instead of analyzing engagement or chart performance, this project focuses on audio characteristics that shape how music feels. These emotional cues provide a clear foundation for later comparing the emotional profiles of songs to the emotional themes expressed in Studio Ghibli film descriptions.
Data Dictionary
To keep this section readable, I included a small preview of the dataset below to show the structure and types of variables used. The full data dictionary, which lists and defines all variables, is linked below for completeness.
Link to full data dictionary
Summary Statistics
To summarize the emotional and sensory characteristics of the music in this dataset, I calculated basic descriptive statistics for a subset of Spotify audio features that are directly related to how a track feels.
valence_% energy_% danceability_% acousticness_%
Min. : 4.00 Min. : 9.00 Min. :23.00 Min. : 0.00
1st Qu.:32.00 1st Qu.:53.00 1st Qu.:57.00 1st Qu.: 6.00
Median :51.00 Median :66.00 Median :69.00 Median :18.00
Mean :51.43 Mean :64.28 Mean :66.97 Mean :27.06
3rd Qu.:70.00 3rd Qu.:77.00 3rd Qu.:78.00 3rd Qu.:43.00
Max. :97.00 Max. :97.00 Max. :96.00 Max. :97.00
liveness_% bpm
Min. : 3.00 Min. : 65.0
1st Qu.:10.00 1st Qu.:100.0
Median :12.00 Median :121.0
Mean :18.21 Mean :122.5
3rd Qu.:24.00 3rd Qu.:140.0
Max. :97.00 Max. :206.0
Descriptive Analysis
To begin exploring the sensory-emotional qualities of the music in this dataset, I focus on a subset of Spotify audio features that describe a track’s overall “feel”: valence, energy, acousticness, loudness, and tempo. These variables allow each song to be represented along measurable emotional dimensions such as brightness, intensity, warmth, and pacing. The following visualizations summarize how these features are distributed across tracks and how they relate to one another. Together, they outline the emotional landscape of the dataset and provide a foundation for later comparing these musical patterns to the emotional themes found in Studio Ghibli film descriptions.
2.1 Valence (Positivity) Distribution
Valence captures how positive or emotionally “bright” a song sounds, making it one of the most direct indicators of musical mood in the dataset. Looking at the distribution of valence scores helps show whether most songs lean cheerful and uplifting, stay closer to neutral, or trend toward darker emotional tones. This plot sets the baseline for understanding the overall emotional atmosphere of the music.
The distribution shows that emotional positivity is spread broadly across the full range rather than clustering at one extreme. Instead of most tracks being overwhelmingly cheerful or somber, the dataset contains a wide mix of moods. There’s some concentration in the mid-range and slightly toward higher valence values, but no single dominant emotional tone emerges. This suggests that the musical landscape represented here is emotionally varied, offering a balance between brighter, upbeat tracks and more neutral or subdued ones.
2.2 Energy (Intensity) Distribution
Energy reflects the overall intensity of a track-whether it feels calm and gentle or loud and driving. Examining the distribution of energy scores helps reveal the emotional temperature of the dataset. This visualization shows whether the collection of songs is mostly high-energy and stimulating, lower-energy and relaxed, or spread across a range of intensities.
Unlike the broad spread seen in valence, the distribution of energy is noticeably concentrated toward the higher end of the scale. Most songs fall roughly between 60%-80% energy, with far fewer tracks in the low-intensity range. This suggests that the dataset is dominated by songs that feel lively and high - energy, while calmer tracks make up a much smaller portion of popular streaming music.
2.3 Acousticness (Warmth/Organic Sound) Distribution
Acousticness measures how natural or organic a track sounds, as opposed to being heavily electronic or produced. This feature is important as it shapes how listeners experience the emotional tone of a song. Acoustic tracks often feel warmer or more intimate, while low-acoustic tracks can feel more polished. Visualizing the distribution helps clarify what kind of sound environment these songs create.
The distribution of acousticness has a strong right tail skew. A large portion of the songs have very little acoustic content, indicating that the dataset is built on electronic or heavily produced sound. Together with earlier plots, this suggests that while the dataset spans a wide range of moods, it does so within a largely modern, high-production sound environment rather than one rooted in acoustic or “unplugged” styles.
2.4 Valence vs. Energy Scatterplot (Emotional Landscape)
By mapping valence against energy, I can visualize the emotional landscape of the dataset in two dimensions. Valence captures how positive or negative a track feels, while energy reflects how intense or mellow it is. This scatterplot shows how brightness and intensity pair up in practice and whether certain emotional zones are more common than others. Together, these dimensions help map how emotional tone and intensity occur in popular music.
The scatterplot shows that songs occupy a wide range of combinations of valence and energy, but some patterns are clear. Energy values remain generally high across the graph, regardless of valence level. This indicates that even songs with lower positivity often maintain a strong sense of intensity. There is also a slight upward trend suggesting that happier songs tend to be somewhat more energetic. Instead of forming tight emotional clusters, points are spread across the plane, suggesting that intensity and positivity mix freely. High-energy tracks appear at nearly every level of valence, reinforcing earlier findings that intensity is a major feature of this music collection, even as mood varies.
2.5 Valence by Mode
Mode is a familiar musical signal often associated with emotional tone, where major keys are traditionally linked with happier sounds and minor keys with sadder ones. By comparing valence across modes, this visualization tests whether those conventional associations hold in this dataset.
This boxplot shows that distribution for major and minor keys overlap substantially. Both major and minor songs span a wide range of valence values, and their median positivity levels differ only slightly. This suggests that, within this dataset, musical mode alone does not strongly determine how positive a song sounds. Instead, emotional tone appears to be shaped more by factors like production choices, energy level, and overall sound design than by key mode itself.
Secondary Data Source
To complement the emotional features of the Spotify dataset, I incorporated a secondary dataset from the Studio Ghibli API. This API provides structured information about each Ghibli film, including the film title, director, release year, and a short plot description. For this project, the plot descriptions serve as the primary analytical component due to containing narrative language that can be examined using sentiment lexicons. These sentiment measures provide a contrast to Spotify’s audio-based emotion indicators.
Using the NRC, Bing, and AFINN lexicons, I extracted emotional patterns from the film descriptions to understand how Ghibli storytelling conveys mood through language. These sentiment summaries reveal the emotional tendencies across films and directors, showing how emotions such as joy, fear, anticipation, sadness, and trust appear within the narrative description.
Emotional Comparison: Spotify Music vs. Ghibli Storytelling
Both datasets reflect emotion, but they do so through different channels. In the Spotify dataset, emotional tone is expressed through sound-based features such as energy, acousticness, and overall production intensity. Valence captures variation in musical brightness rather than a clear positive-negative divide. In contrast, Studio Ghibli descriptions convey emotion through narrative language tied to themes, atmosphere, and story progression. Comparing these two approaches highlights how emotional expression depends not just on what is being felt, but on the medium through which it is communicated.
Range of Emotion
Spotify songs cover a wide span of valence, varying from bright to dark moods. This emotional variation, however, occurs within a consistent sonic context. Most tracks remain high in energy and low in acousticness, indicating that even when emotional tone shifts, the overall intensity and production style stays similar. While songs may differ in how bright or subdued, they feel, they tend to deliver that emotion through the same high- energy, digitally produced sound environment.
Ghibli descriptions show a different pattern. Emotional language is distributed across multiple categories, with no single emotion dominating. Rather than emphasizing one emotional dimension, the descriptions suggest stories where multiple feelings appear together and change over time as the narrative unfolds.
Intensity and Structure
In the Spotify dataset, emotional intensity remains relatively consistent even as musical positivity varies. Songs span a wide range of valence values, but most maintain high energy levels regardless of whether they sound brighter or darker. This indicates that emotional tone in popular music often shifts in mood without a corresponding change in intensity.
Studio Ghibli film descriptions show a different emotional structure. Emotional language reflects recognizable patterns shaped by storytelling choices, with certain directors emphasizing anticipation and joy while others lean more toward sadness or trust. These patterns suggest that emotional tone in Ghibli films is guided by narrative arc and context, allowing emotions to build, shift, and resolve over time rather than remaining fixed in intensity.
Sound vs. Story
Spotify’s low acousticness and consistently high energy create an emotional environment driven by production, clarity, and momentum. The emotional effect is immediate and sensory, the feeling happens in the moment you hear it. Ghibli’s emotional profile, built from descriptive language, reflects shifting themes and atmosphere. Tenderness, conflict, hope, loss, and wonder appear not as isolated states but as parts of a developing world.
Why Comparison Matters:
Together, these patterns show that emotion behaves differently across mediums. Music conveys feeling through intensity and sonic qualities, delivering emotion more at once. Ghibli storytelling conveys feeling throughout context and progression, allowing multiple emotions to accumulate and transform. Comparing the two highlights that emotional expression is not singular – it adapts depending on whether we experience it through sound or narrative. This makes examining them side by side worthwhile: it reveals two complementary ways that creative media shape how we understand and feel emotion.
