madol-abraham/Agricultural_chatbot

This the chatbot specifically design to response to the farmers in south sudan. # Agricultural Chatbot for South Sudan This project implements a domain-specific chatbot focused on agricultural advice for farmers in South Sudan. The chatbot uses natural language processing techniques to provide relevant responses to farming-related questions. Below are the screenshot of the chatbot: ## Project Overview The chatbot is designed to assist farmers with agricultural information specific to South Sudan's climate and farming conditions. It can answer questions about: - Crop selection and planting times - Soil fertility management - Pest and disease control - Water management - Post-harvest techniques - Market access ## Dataset The dataset consists of 300 question-answer pairs related to agriculture in South Sudan. Each pair includes a farming question and a corresponding expert answer. The dataset is structured to cover various aspects of farming relevant to the region. Sample data: ``` Question: "What are the best crops to grow in South Sudan during the rainy season?" Answer: "Maize, sorghum, groundnuts, and sesame are good crops for the rainy season in South Sudan." ``` ## Data Preprocessing The preprocessing pipeline includes: 1. Text normalization (converting to lowercase) 2. Special character removal 3. Tokenization using NLTK 4. Stopword removal 5. Lemmatization This ensures the text is clean and standardized before being processed by the model. ## Model Architecture The chatbot uses a pattern matching approach with similarity scoring to find the most relevant answers to user queries. This approach was chosen for its reliability and efficiency in domain-specific applications. ## Hyperparameter Tuning Extensive hyperparameter tuning was performed to optimize the model's performance: | Learning Rate | Batch Size | Epochs | Validation Loss | Perplexity | Improvement (%) | |---------------|------------|--------|----------------|------------|-----------------| | 5e-5 | 8 | 3 | 1.2134 | 3.3651 …