Is there a way to add REGEX pipeline components to the SimpleKGPipeline?
I have unstructured documents with annotations in the text, conforming to a strict taxonomy, e.g. “sxxxxxx” for a section, “ixxxxxx” for an image. “xxxxxx” being numeric.
Looks as if I may have to write a Custom component and switch to using a custom pipeline ?
Hey,
If this component is meant to replace (or extend) the entity and relation extraction component, then yes, writing a custom component and pipeline is the only available option for now.
Yes, to augment. Is there any information on writing such a component? Derive from the llm extractor component and don’t invoke an llm?
You can find an example here: neo4j-graphrag-python/examples/customize/build_graph/components/extractors/custom_extractor.py at main · neo4j/neo4j-graphrag-python · GitHub
The package has a base EntityRelationExtractor you can subclass.