The goal of this project is to use Natural Language Processing (NLP) to extract insights from text data, specifically by conducting sentiment analysis and generating visualizations through word clouds. The main objective is to perform an in-depth analysis of the song lyrics of "Nightstalker", a Greek Stoner-Rock band formed in 1989 and one of my personal favorites. More information about the band can be found on their official page and wikipedia.
This exploratory portfolio project applies text preprocessing, word-frequency analysis, word clouds, and lexicon-based sentiment scoring to a snapshot of English-language lyrics by Nightstalker, a Greek heavy-rock band formed in 1989.
The dataset used in the original analysis contained 73 songs across eight releases. The purpose is to demonstrate a basic natural-language-processing workflow on cultural text, not to determine the band's intentions or objectively classify the emotions of its music.
- Which words appear most frequently after basic text cleaning?
- How do word-frequency patterns vary across releases?
- How do VADER and TextBlob score the available lyrics?
- How do the resulting scores vary across songs and albums?
- Compile the lyric references used for the project.
- Normalize case and remove punctuation and selected stopwords.
- Tokenize the cleaned text.
- calculate word frequencies and generate word clouds.
- Apply VADER and TextBlob sentiment scoring.
- Aggregate and visualize the scores by song and album.
- Python
- pandas
- NLTK
- VADER
- TextBlob
- Matplotlib
- WordCloud
- Jupyter Notebook
- Cleaned-token and word-frequency summaries
- Word clouds
- Song-level and album-level sentiment-score comparisons
- Supporting visualizations
- Clone or download the repository.
- Create a virtual environment.
- Install the packages listed in
requirements.txt. - Supply a lawfully obtained text corpus in the documented input format.
- Run the notebook cells in order.
- VADER and TextBlob are general-purpose, lexicon-based tools; they are not music- or lyrics-specific emotion models.
- Metaphor, irony, narrative voice, negation, and song context can produce misleading sentiment scores.
- Word clouds show frequency, not thematic importance.
- Album comparisons may be influenced by unequal track counts and lyric lengths.
- Results should be described as model-generated sentiment scores, not as the band's verified emotions or intentions.
In the medium article , you'll get an in-depth look at the methodologies I used, the objectives I set out to achieve, and the exciting discoveries I made along the way.
Feel free to reach out if you have any questions or thoughts to share!