Personal project with three text mining cases on Indonesian text, all in impl/Text_Mining_Experiment.ipynb.
| Case | Data | Approach |
|---|---|---|
| Hate Speech Classification | data_1A.csv |
Sentence-BERT embeddings (firqaaa/indo-sentence-bert-base) + MLP tuned with Optuna, per-label thresholds |
| Persona Analysis of Tweets | data_2A.csv |
TF-IDF + LSA + K-Means (k chosen by silhouette score), LDA topics |
| Text Summarization | data_3A.csv |
Extractive TextRank, sentence count chosen by ROUGE-L |
Each case ends with an inference example.
data/ input CSVs
impl/ notebook
result/ plots and CSV outputs (created on run)
result/model/ best model of each case
result/cache/ cached sentence embeddings
pip install numpy pandas matplotlib nltk scikit-learn keras tensorflow optuna sentence-transformers rouge wordcloud networkx joblib
Run the notebook from impl/, starting with the Shared Setup cell. Runs on CPU; Case 1 is the slowest (first run downloads the embedding model).