GROBID is a machine learning library for extracting, parsing, and re-structuring raw documents. It is designed and expected to be used to parse academic papers, where it works particularly well. Note: if the articles supplied to Grobid are large documents (e.g. dissertations) exceeding a certain number of elements, they might not be processed. This page covers how to use the Grobid to parse articles for LangChain.

​
Installation

The grobid installation is described in details in https://grobid.readthedocs.io/en/latest/Install-Grobid/. However, it is probably easier and less troublesome to run grobid through a docker container, as documented here.

​
Use Grobid with LangChain

Once grobid is installed and up and running (you can check by accessing it http://localhost:8070), you’re ready to go. You can now use the GrobidParser to produce documents 
from langchain_community.document_loaders.parsers import GrobidParser
from langchain_community.document_loaders.generic import GenericLoader

#Produce chunks from article paragraphs
loader = GenericLoader.from_filesystem(
    "/Users/31treehaus/Desktop/Papers/",
    glob="*",
    suffixes=[".pdf"],
    parser= GrobidParser(segment_sentences=False)
)
docs = loader.load()

#Produce chunks from article sentences
loader = GenericLoader.from_filesystem(
    "/Users/31treehaus/Desktop/Papers/",
    glob="*",
    suffixes=[".pdf"],
    parser= GrobidParser(segment_sentences=True)
)
docs = loader.load()
Chunk metadata will include Bounding Boxes. Although these are a bit funky to parse, they are explained in https://grobid.readthedocs.io/en/latest/Coordinates-in-PDF/