One of the main products of the Campus Analytics project is what we call the EPFL Knowledge Graph—a database of teaching, research, and entrepreneurial content, all interconnected in a unified network of academic knowledge.
On top of this graph, a number machine learning algorithms continuously analyse the unified data, producing statistical insights, choice recommendations, and answers to complex institutional questions.
There are currently two main applications that make use of the EPFL Knowledge Graph: the Graph Search web interface and the Graph Engine application programming interface (API).
The Graph Search web interface (available for beta testing at graphsearch.epfl.ch) is a website where students and researchers can search for topics they wish to learn and obtain recommendations on, among many things, which courses to attend or which publications to read. It provides sophisticated AI-powered features such as:
- searching for concepts inside video lectures, and instantly navigating to the timestamps where concepts are addressed,
- getting course recommendations based on content similarity and student performance,
- recommending paper co-authorships based on EPFL’s research network,
- searching for tech-transfer and startup investment opportunities.
The Graph Engine API is a more powerful, general purpose, graph processing engine that can be used by EPFL’s research labs and administrative teams to directly query the knowledge graph and obtain answers to complex questions. It provides features such as:
- a native graph querying language,
- data subset extraction and exports,
- out-of-the-box machine learning, natural language processing, and graph processing operations on the knowledge graph.
The EPFL Knowledge Graph is partially open for research use. If you’re a student interested in data engineering and machine learning, don’t hesitate to consult our list of semester projects. If you’re a PhD candidate, postdoc, or administrator interested in collaborating with us, don’t hesitate to contact us.