SmartSearch (AI Enhanced Multi-Modal Search Engine)
Project Description
Motivation:
When you want to search some content in medical,
google only support text-only input or image-only input.
you may not express what you want to search clearly and correctly.
hard to find reliable material.
What this system would be look like:
Medbook and Website Database: The knowledge database we want to query from.
Multi-Modal RAG strategy: Accurate retrieval.
LLM & MLLM: Understand what you want to search and verify the retrieved samples.
Why should we do this?
It is a time to incoporate AI in search as the understanding and reasoning capabilities of the LLM.
We have such database and such model.
The difficulties lie in:
Advanced Multi-Modal RAG strategy.
Construction of an effective search engine.
When you want to search some content in medical,
google only support text-only input or image-only input.
you may not express what you want to search clearly and correctly.
hard to find reliable material.
What this system would be look like:
Medbook and Website Database: The knowledge database we want to query from.
Multi-Modal RAG strategy: Accurate retrieval.
LLM & MLLM: Understand what you want to search and verify the retrieved samples.
Why should we do this?
It is a time to incoporate AI in search as the understanding and reasoning capabilities of the LLM.
We have such database and such model.
The difficulties lie in:
Advanced Multi-Modal RAG strategy.
Construction of an effective search engine.
Supervisor
CHEN, Hao
Quota
5
Course type
UROP1000
UROP1100
UROP2100
UROP3100
UROP3200
UROP4100
Applicant's Roles
Deep learning, Image Processing
Applicant's Learning Objectives
Potentially publish top-tier papers
Complexity of the project
Moderate