Climi/Climate-Education-QA-Chatbot

Generative QA Chatbot for Climate Education

This chatbot help users (especially students, young activists, or the general public) learn about climate change, its causes, impacts, solutions, and key concepts through conversational Q&A.

• Model: T5-small (Text-To-Text Transfer Transformer)
• Framework: TensorFlow
• Evaluation Metrics: BLEU Score

Domain Justification:

Climate education chatbots address the critical need for accessible, accurate climate science information. Think of it like having a climate science teacher available 24/7 who can explain complex concepts like carbon cycles, greenhouse effects, or climate policies in simple terms.

Architecture Breakdown:

• Architecture Type: Encoder-Decoder Transformer
• Layers: 6 Encoder + 6 Decoder
• Parameters: 60,506,624 (60M)
• Size: ~240 MB
• Performance: 0.0549 BLEU, ~17s generation
• Attention Mechanism: Multi-Head Self-Attention
• Position Encoding: Relative Position Bias
• Activation Function: ReLU

Checkout the deployed QA chatbot on Streamlit Cloud: https://ayika-app-v1.streamlit.app/

Author: Eunice Adewusi Climiradi

My Links: https://linktr.ee/climiradi

Date: June 2025