Mealera: Nigerian Food & Wellness Conversation Model
Mealera is a state-of-the-art conversational AI model designed to understand, classify, and respond to queries about Nigerian food, meal planning, health, shopping, and cultural dietary needs. Built on DistilBERT and fine-tuned on thousands of real-world, culturally rich conversations, Mealera empowers digital food platforms, health apps, and smart assistants to deliver context-aware, locally relevant, and health-conscious experiences for users in Nigeria and beyond.
1import torch
2query ="What can I cook for a family of four with 2000 naira?"3inputs = tokenizer(query, return_tensors="pt", truncation=True, padding=True)4with torch.no_grad():5 outputs = model(**inputs)6 logits = outputs.logits
7 predicted_class_id = logits.argmax().item()8print("Predicted context:", label_encoder.inverse_transform([predicted_class_id])[0])
🏷️ Context Labels (Intents)
The model predicts one of several context labels, including:
recipe_recommendation: Suggesting recipes or meal ideas
casual_chat: General conversation or greetings
greeting: Salutations and opening messages
vendor_recommendation: Finding food vendors or markets
budget_meal: Affordable meal suggestions
health_advice: Nutrition and health-related queries
dietary_restrictions: Special diets (e.g., allergies, intolerances)
shopping_list: Generating shopping lists
meal_plan: Weekly/daily meal planning
allergy_concern: Allergy and food intolerance questions
(See label_encoder.pkl for the full list and mapping.)
📦 Training & Evaluation
Base Model: distilbert-base-uncased
Training Data: 2,248+ annotated Nigerian food and wellness conversations
