06. Word

Microsoft Word

Microsoft Word is a word processor developed by Microsoft.

This covers how to load a Word document into a document format that can be used downstream.

Docx2txtLoader

You can use docx2txt to import .docx files into documents.

Copy

# installation
# !pip install -qU docx2txt

Copy

from langchain_community.document_loaders import Docx2txtLoader

loader = Docx2txtLoader("./data/sample-word-document.docx")  # Initialize document loader

docs = loader.load()  # loading documents

print(len(docs))

Copy

1

UnstructuredWordDocumentLoader

Copy

from langchain_community.document_loaders import UnstructuredWordDocumentLoader

# Output of the uploaded document
loader = UnstructuredWordDocumentLoader("./data/sample-word-document.docx")

# data loading
docs = loader.load()

print(len(docs))

Copy

1

The result is loaded as a single Document.

Copy

# metadata output of power
print(docs[0].metadata)

Copy

{'source': './data/sample-word-document.docx', 'filename': 'sample-word-document.docx', 'file_directory': './data', 'last_modified': '2024-07-30T03:09:26', 'filetype': 'application/vnd.openxmlformats-officedocument.wordprocessingml.document', 'category_depth': 0, 'page_number': 1, 'category': 'Title'}

Internally, amorphism creates different “elements” for each chunk of text.

By default these are combined together, but can be easily separated by specifying mode="elements" .

Copy

# UnstructuredWordDocumentLoader
loader = UnstructuredWordDocumentLoader(
    "./data/sample-word-document.docx", mode="elements"
)

# data loading
docs = loader.load()

# Output of the uploaded document
print(len(docs))

Copy

125

Copy

# Print the contents of the first document
print(docs[0].page_content)

Copy

Semantic Search

Copy

# Print the contents of the first document
docs[0].metadata

Copy

{'source': './data/sample-word-document.docx', 'filename': 'sample-word-document.docx', 'file_directory': './data', 'last_modified': '2024-07-30T03:09:26', 'filetype': 'application/vnd.openxmlformats-officedocument.wordprocessingml.document', 'category_depth': 0, 'page_number': 1, 'category': 'Title'}