amandatheuri/NLP-for-English-Sheng-Swahili

This project is a Natural Language Processing (NLP) application that automatically identifies the language of short text inputs. # Language Identification System (NLP Project) ## Overview This project is a Natural Language Processing (NLP) application that automatically identifies the language of short text inputs. It supports multiple languages commonly used in Kenya: * English * Swahili * Sheng * Kikuyu The system was developed as part of an academic project to demonstrate text classification using machine learning techniques. --- ## Objectives * Build a model that can classify short text into different languages * Apply NLP preprocessing and feature extraction techniques * Compare multiple machine learning models * Deploy the model as an interactive application --- ## Technologies Used * Python * Scikit-learn * Pandas / NumPy * Streamlit --- ## Methodology ### 1. Data Collection A balanced dataset was created with text samples from four languages: * English * Swahili * Sheng * Kikuyu Each class contains an equal number of samples. --- ### 2. Data Preprocessing Text data was cleaned using: * Lowercasing * Removal of extra whitespace * Preservation of special characters (important for Kikuyu) --- ### 3. Feature Extraction Text was converted into numerical features using: * TF-IDF Vectorization * Character n-grams (1–5 range) This approach captures language-specific character patterns. --- ### 4. Model Training Two models were trained and compared: * Multinomial Naive Bayes * Logistic Regression --- ### 5. Model Evaluation Performance was evaluated using: * Accuracy * Precision * Recall * F1-score * Confusion Matrix #### Results: * Naive Bayes Accuracy: 100% * Logistic Regression Accuracy: 97.92% Naive Bayes was selected as the final model due to its superior performance. --- ### 6. Deployment The model was deployed using Streamlit, allowing users to: * Enter text * Click a button * Receive an instant language prediction --- ## How to Run the Project ### 1. Clone the repository ```bash git clone github.com …