With the explosion of image models and embedding options for multimedia (including PDFs!), we should keep track of various tools and strategies for a few reasons. First so that we can direct users on how they might wrap some of these resources as Tika parsers to get the benefits of the Tika framework (esp. on embedded documents), and second, so that if there are a few emerging leaders/techniques, we might consider adding wrappers to the main Tika project.
General Info Blog Posts
"How to" Blog Posts
- Demo using https://unstructured.io/ 's
partition_pdf
(): high-precision-rag-for-table-heavy-documents-using-langchain-unstructured-io-kdb-ai-22f7830eac9a - unstructured.io and Mistral: https://medium.com/django-unleashed/leveraging-unstructured-io-and-mistralai-for-advanced-pdf-data-processing-331c2c3683fb
- Demo using ColPali (Vision Language Model) to create embeddings from rendered PDF pages: https://blog.vespa.ai/the-rise-of-vision-driven-document-retrieval-for-rag/
Conference Talks
Conferences
Open Source Tools (no particular order/not an endorsement)
- https://github.com/Unstructured-IO/unstructured (as of this writing, uses poppler, tesseract, libreoffice and pandoc)
Select Open Source Tools that haven't been updated in the last year (no particular order/not an endorsement)
Commercial Tools (no particular order/not an endorsement)
Non-FAANGM Companies Focusing on Advanced Document Processing (no particular order/not an endorsement)
Cloud Services (no particular order/not an endorsement)
- https://azure.microsoft.com/en-us/products/ai-services/ai-document-intelligence
- https://aws.amazon.com/solutions/ai/intelligent-document-processing/
- https://cloud.google.com/document-ai
Research Papers
- Good luck. There are too many...unless you have a department of grad students at the ready