reasoning-augmented-generation

Reasoning Augmented Generation - BH 2024

Triple extraction

cd src
python pipeline.py "proteins that are involved the process of apoptosis"

Output

The extracted triple is: (bcl-2, involved, apoptosis)

Detected role:  involved -> part of
Detected class:  apoptosis -> Apoptosis modulation

OWL sentence: part of some Apoptosis modulation

Schema extraction

Given a query in natural language and the schema of the SPARQL endpoint (for example Uniprot), we retrieve the schema of the query by interacting with the LLM in several steps:

• Class Extraction
• Predicate Extraction
• Query Generation

Example

We show an example for the query: List all Human UniProt entries and their computationaly potential isoforms.