DeepL Document Transformer

[j-tara]

Checked

• I searched existing ideas and did not find a similar one
• I added a very descriptive title
• I've clearly described the feature request and motivation for it

Feature request

We would like to add a Document Transformer that uses DeepL to enhance writing and translations of text and documents:

• Perform high-quality translations of text in 33 languages
• Support customizable translations including the glossary feature, which allows users to define domain-specific vocabulary to ensure consistency and accuracy in specialized translations (ie. medical or legal terminology)
• Perform translations on files (pdf, docx, xslx, pptx, txt, html, xliff), useful for the efficient batch processing of different file types
• Support customizable written output with choice of formality and tone levels

Motivation

We noticed that Langchain does not currently have an integration with DeepL. DeepL is a powerful Language AI platform, specializing in enhanced writing and high quality language translations through its products DeepL Translator and DeepL Write.

It is highly suitable as a document transformer due to its existing compatibility with several file types and customizable output. We believe that with a DeepL integration, developers can easily set up high-quality and customizable document processing pipelines.

Proposal (If applicable)

We plan to use the DeepL API in our implementation.

Files to modify:

• libs/langchain/langchain/document_transformers/__init__.py: to register the DeepL document transformer in the main Langchain library
• libs/community/langchain_community/document_transformers/__init__.py: to register DeepL document transformer in the community section
• libs/langchain/tests/unit_tests/document_transformers/test_imports.py: to import unit tests in the main Langchain library
• libs/community/tests/unit_tests/document_transformers/test_imports.py: to import unit tests in the community section

Files to create:

• libs/langchain/langchain/document_transformers/deepl.py: to implement the DeepL document transformer in the main Langchain library
• libs/community/langchain_community/document_transformers/deepl.py : to implement the DeepL document transformer in the community section
• docs/docs/integrations/document_transformers/deepl.ipynb: to provide documentation and examples on how to use the DeepL document transformer

We are a few students from the University of Toronto, aiming to develop this feature and submit a pull request by early November.