| # NLP Sentiment Analysis Project | |
| <img src="https://i.ibb.co/N27JpMw0/512-1x-shots-so.png" alt="512 1x shots so" border="0"> | |
| ## Project Overview | |
| This project builds a text classification pipeline for emotion detection using NLP techniques. The workflow includes data loading, preprocessing, feature extraction, model training, and evaluation. | |
| ## What Was Done | |
| - Loaded the dataset from `train.txt`. | |
| - Cleaned and preprocessed text data. | |
| - Converted emotion labels to numeric values. | |
| - Vectorized text using Bag-of-Words and TF-IDF. | |
| - Trained and evaluated multiple classification models. | |
| ## Data Files | |
| - `train.txt`: Training dataset containing text and emotion labels. | |
| - `test.txt`: Optional test data file for further evaluation. | |
| - `val.txt`: Optional validation data file for model tuning. | |
| ## Preprocessing Steps | |
| - Lowercased all text. | |
| - Removed URLs. | |
| - Removed digits. | |
| - Removed emojis. | |
| - Removed punctuation. | |
| - Removed stop words. | |
| ## Models Evaluated | |
| - Multinomial Naive Bayes | |
| - Logistic Regression | |
| - Support Vector Machine (SVM) | |
| ## Results | |
| The models were evaluated using: | |
| - Accuracy | |
| - Precision | |
| - Recall | |
| ## Summary Table | |
| | Step | Description | | |
| | -------------- | -------------------------------------------------------------------- | | |
| | Data Loading | Read `train.txt` with text and emotion labels. | | |
| | Label Encoding | Mapped emotion labels to integer values. | | |
| | Text Cleaning | Lowercase, remove URLs, digits, emojis, punctuation, and stop words. | | |
| | Vectorization | Converted text into numeric features using Bag-of-Words and TF-IDF. | | |
| | Model Training | Trained Naive Bayes, Logistic Regression, and SVM models. | | |
| | Evaluation | Compared models using accuracy, precision, and recall. | | |
| ## Notes | |
| This project focuses on text preprocessing and classification for emotion detection. The notebook includes the full pipeline for preparing the data and evaluating models. | |