NVIDIA Ingest - NVIDIA's open-source intelligent document extraction and structuring tool
NVIDIA Ingest is an open-source collection of microservices from NVIDIA for parsing complex, messy, unstructured PDFs and other enterprise documents. NVIDIA Ingest converts documents into metadata and text, making them easy to embed into retrieval systems...
What is NVIDIA-Ingest?
NVIDIA Ingest is an open-source collection of microservices from NVIDIA for parsing complex, messy, unstructured PDFs and other enterprise documents. NVIDIA Ingest converts documents into metadata and text, facilitating embedding into retrieval systems. It supports multiple document formats, including PDF, Word, PowerPoint, and images, and offers various extraction methods to balance throughput and accuracy. NVIDIA Ingest supports preprocessing and post-processing operations such as text segmentation, transformation, filtering, embedding generation, and image storage. Based on parallel document processing, NVIDIA Ingest improves extraction efficiency and supports embedding extracted content into vector databases such as Milvus, making it suitable for large-scale document processing and generative applications.
Key features of NVIDIA-Ingest
- Multi-format document supportIt supports parsing various complex enterprise document formats such as PDF, Word (Docx), PowerPoint (Pptx), and images.
- Multiple extraction methodsIt supports multiple extraction methods, facilitating a trade-off between throughput and accuracy. For example, PDF documents can be extracted using pdfium, Unstructured.io, and Adobe Content Extraction Services.
- Content Classification and ExtractionThe document content is categorized into text, tables, charts, and images, and each category is extracted. Optical Character Recognition (OCR) technology is used to further contextualize the extracted content and convert it into a well-defined JSON format.
- Parallel processingIt supports splitting documents into pages and extracting content from each page in parallel, improving processing efficiency.
- Preprocessing and postprocessingIt supports a variety of preprocessing and postprocessing operations, including text segmentation and chunking, content transformation, filtering, embedding generation, and image storage.
The technical principles of NVIDIA-Ingest
- microservice architectureBased on a microservice architecture, each microservice is responsible for specific processing tasks, such as text extraction, image extraction, and table extraction, thereby improving the scalability and flexibility of the system.
- GPU accelerationBased on NVIDIA's GPU technology, especially the H100 and A100 GPUs, it accelerates the document parsing and content extraction process. The parallel computing capabilities of the GPU significantly improve processing efficiency, especially when processing large numbers of documents.
- Optical Character Recognition (OCR)Uses OCR technology to convert images and tables in documents into readable text. NVIDIA-Ingest integrates multiple OCR engines, such as PaddleOCR, to improve the accuracy and efficiency of text recognition.
NVIDIA-Ingest project address
- GitHub repository:https://github.com/NVIDIA/nv-ingest
Application scenarios of NVIDIA-Ingest
- Enterprise Content ManagementConvert paper documents, PDFs, Word documents, and PowerPoint presentations into searchable and editable digital formats to support knowledge sharing and collaboration.
- Intelligent Customer Service SystemIt parses user-uploaded documents, extracts key information, generates automatic responses, and improves customer service efficiency and user satisfaction.
- Legal and compliance fieldsIt analyzes contracts and legal documents, extracts key terms and conditions, and supports compliance checks, risk assessments, and case management.
- Financial industryIt analyzes financial reports, contracts, and market research documents to extract key data, supporting risk assessment, compliance monitoring, and customer due diligence.
- healthcareIt converts medical records into structured data, supporting electronic medical record management, clinical research, and medical image analysis.