MBTI Classifier is an educational Natural Language Processing (NLP) project that predicts MBTI-style personality traits from user-provided text. The project uses text preprocessing, bag-of-words style features, and Naive Bayes classifiers.
This project is designed as a machine learning portfolio demo. It should not be treated as an official psychological assessment.
Try the Streamlit demo here:
https://mbti-classifier-nlp-elkew8jedtccsxd9exyxoh.streamlit.app/
Disclaimer: This demo is for educational and portfolio purposes only. It is not an official psychological test, diagnosis tool, or personality assessment.
The goals of this project are to demonstrate:
- basic NLP text preprocessing,
- text classification workflow,
- bag-of-words feature extraction,
- Naive Bayes baseline classification,
- simple model evaluation,
- Streamlit demo deployment,
- responsible communication of machine learning limitations.
- User text input
- Text preprocessing
- MBTI-style trait prediction
- Naive Bayes classification
- Trait-level confidence display if available
- Streamlit web interface
- Educational disclaimer
- Python
- pandas
- NumPy
- NLTK
- Naive Bayes Classifier
- Streamlit
- Jupyter Notebook
mbti-classifier-nlp/
├── app.py
├── requirements.txt
├── README.md
├── mbti_classifier_notebook.ipynb
├── data/
│ └── dataset.zip
└── screenshots/
├── 01-streamlit-home.png
├── 02-text-input.png
└── 03-prediction-result.png
The dataset is stored in:
data/dataset.zip
The CSV inside the ZIP should contain at least these columns:
type
posts
Column description:
| Column | Description |
|---|---|
type |
MBTI label such as INFP, INFJ, INTP, ENFP, etc. |
posts |
Text posts used as input data for NLP classification |
For privacy and safety reasons, the public demo should not expose raw personal text samples.
The text is cleaned using steps such as:
- converting text to lowercase,
- removing URLs,
- removing punctuation and non-alphabetic characters,
- removing stopwords,
- extracting useful word tokens.
The cleaned text is converted into simple word-based features using a bag-of-words style representation.
The project uses Naive Bayes classifiers as baseline models.
The demo predicts MBTI-style traits through four binary classifiers:
- Introvert vs Extrovert
- Intuition vs Sensing
- Thinking vs Feeling
- Judging vs Perceiving
The predicted letters are combined into one MBTI-style output.
Example output:
INFJ
INTP
ENFP
git clone https://github.com/frnqpur/mbti-classifier-nlp.git
cd mbti-classifier-nlpWindows:
python -m venv .venv
.venv\Scripts\activateMac/Linux:
python3 -m venv .venv
source .venv/bin/activatepip install -r requirements.txtMake sure the dataset exists at:
data/dataset.zip
streamlit run app.pyThen open:
http://localhost:8501
This project is deployed using Streamlit Community Cloud.
Recommended deployment settings:
Repository: frnqpur/mbti-classifier-nlp
Branch: main
Main file path: app.py
cPanel is not recommended for this project because many shared hosting plans do not support Python machine learning dependencies such as pandas and NLTK.
Enter an anonymous sample text, for example:
I enjoy learning new ideas, reflecting on future possibilities, and writing about personal growth. I prefer meaningful conversations and structured plans.
The app will display:
- predicted MBTI-style type,
- trait-level predictions,
- confidence values if available,
- educational disclaimer.
This project is for educational and portfolio purposes only.
The prediction result is not:
- an official psychological test,
- a diagnosis,
- a counseling tool,
- a hiring assessment,
- a formal personality evaluation.
The model output should not be used for medical, psychological, hiring, counseling, or other sensitive decisions.
This project has several limitations:
- The model is a baseline model.
- Feature extraction is simple.
- Dataset may contain bias.
- Prediction depends on input text quality and length.
- Confidence values do not represent psychological validity.
- The system is not production-ready.
- The result should not be interpreted as an official MBTI assessment.
Possible improvements include:
- using TF-IDF feature extraction,
- trying Logistic Regression or Linear SVM,
- adding cross-validation,
- adding classification reports,
- saving trained models using pickle or joblib,
- improving Streamlit UI,
- adding anonymous sample inputs,
- creating a model card,
- improving dataset privacy handling.
Built an educational NLP text classification demo using Python, text preprocessing, bag-of-words features, and Naive Bayes classifiers to predict MBTI-style traits from user input. The project includes a Streamlit demo and clear limitations to avoid presenting the output as an official psychological result.
Use an appropriate license based on the dataset source and project ownership.