Repository navigation
This is a robust and scalable pipeline built to transform raw scholarly abstracts into a fully enriched knowledge graph, which is then ingested into a Neo4j database. The system utilizes asynchronous processing and LLM-based entity extraction, coupled with text embedding and advanced graph construction techniques. These features ensure efficient transformation and ingestion of research data, all of which are configurable without the need for code changes.
As part of this pipeline, an example of the Neo4j knowledge graph is provided in .json format. This graph is derived by extracting relevant entities and linking them with relationships from 8358 scientific abstracts in the domain of nuclear fusion energy.
HINT: The embeddings was not included in the node properties to keep the size of the file within 2GB limits.