MealMatch

If you want to find food or your meal suitable for your budget so you need to known my platform MealMatch. It can help you figure out

Comment

Inspiration

“What should we eat?”

A simple question… but somehow it makes a whole group think like they’re solving advanced math. One person says “anything is fine” (but rejects everything), another has a 50k budget but wants a 5-star experience.

Meanwhile, current apps just show a long list of restaurants — the decision is still… entirely up to you.

So we thought: What if there was a system that actually understands you — from your budget to your taste — and makes the decision easier?

That’s how MealMatch was born — a “food companion” that never says “whatever you want.”

What it does

MealMatch helps users by allowing them to input:

budget taste description (free-form, e.g., “spicy, soup-based, under 70k”) number of people, dining time

The system will then:

analyze needs using AI extract dining preferences fetch restaurant data from Google Maps analyze reviews to identify popular dishes recommend the most suitable restaurant and specific dishes Key highlights: Understands natural language Recommends specific dishes, not just restaurants Provides explanations for recommendations (Explainable AI)

How we built it

We combined AI with a rule-based system to ensure both intelligence and reliability.

  1. Data sources

Using Google Maps API:

restaurant names ratings price levels reviews

  1. Dish extraction from reviews

We use:

keyword matching (fast fallback) optional AI to extract dish names from reviews

This allows the system to understand what a restaurant serves without needing an official menu.

  1. User input analysis

AI processes natural language and converts it into structured data:

taste (spicy, sweet, etc.) category (soup, dry dishes, etc.) budget context (dinner, group dining, etc.)

  1. Matching & Scoring engine

We match:

user needs dish tags

Then compute a compatibility score.

  1. Recommendation output

The system returns:

restaurant suitable dishes estimated price match percentage AI-generated explanation

Challenges we ran into

  1. No structured menu data

Google Maps does not provide menus, so we had to:

extract from reviews infer from unstructured data

  1. Understanding natural language

User inputs are often vague:

“something warm” “light but still filling”

We needed AI to convert these into structured data.

  1. Balancing AI and accuracy AI is powerful but hard to control Rule-based systems are stable but less flexible

Solution: a hybrid system

  1. Limited development time

We had to:

optimize scope (MVP) focus on core features instead of complex UI

Accomplishments that we're proud of

We successfully built a system that:

understands natural language input recommends real dishes from review data provides explainable recommendations

We effectively combined:

AI (NLP) real-world data (Google Maps) scoring system

Creating an experience similar to a personal food assistant

What we learned

AI is not always the perfect solution → it needs to be combined with traditional logic

Real-world data is often:

messy incomplete → requires smart processing

UX is extremely important Users prefer:

clear recommendations explanations simplicity

Scope limitation is critical when working under tight deadlines

What's next for MealMatch

In the future, we plan to develop:

  1. Personalization learn from eating history understand long-term preferences
  2. Location integration recommend nearby restaurants optimize travel time
  3. Conversational chatbot

Example:

“I feel tired today, what should I eat?”

  1. Advanced recommendation use embeddings / vector search deeper semantic understanding (beyond keywords)
  2. Real menu data crawl data or partner with restaurants

Built With

Share this project:

Updates