Word Documents

This covers how to load Word documents into a document format that we can use downstream.

from langchain.document_loaders import UnstructuredWordDocumentLoader

loader = UnstructuredWordDocumentLoader("example_data/fake.docx")

data = loader.load()

data

[Document(page_content='Lorem ipsum dolor sit amet.', lookup_str='', metadata={'source': 'fake.docx'}, lookup_index=0)]

Retain Elements

Under the hood, Unstructured creates different “elements” for different chunks of text. By default we combine those together, but you can easily keep that separation by specifying mode="elements".

loader = UnstructuredWordDocumentLoader("example_data/fake.docx", mode="elements")

data = loader.load()

data[0]

Document(page_content='Lorem ipsum dolor sit amet.', lookup_str='', metadata={'source': 'fake.docx', 'filename': 'fake.docx', 'category': 'Title'}, lookup_index=0)