konahart/relation-extraction-pipeline

Training pipeline for a relation extraction system

README.md

Hi, I'm Laurel Goodhart! I'm a software engineer who specializes in Natural Language Processing. I have spent my career bringing interactive character experiences to life and making tools for creative people.

Want me on your team? Reach out!

relation-extraction-pipeline's Issues

GigawordCorpusHandler gives incorrect spans

This has to do with the way SAX calls characters(). Still looking for a workaround to provide accurate document spans.

NLP annotation done twice when using CoreNLPProcessor

This is because tokenization, sentence splitting, lemmatization, and POS tagging are all needed to do NER tagging, which is used as a basic way to find entities in sentences, but those annotations basically get thrown away as part of the price of modularity. Stanford CoreNLP can write out XML of its annotations (which is provided as an option here, but produces a lot of output files), but there is currently no way to deserialize the XML and therefore reuse the annotations.

README.md images

Add walkthrough

Maybe website?

Add demo data

Scripts missing

Need to be modified/generalized for usage