Our technology
TEKLIA's artificial intelligence projects rely on a modular processing chain capable of integrating the best available AI models for each document task. Arkindex orchestrates these processes, ensures their traceability, and facilitates their large-scale deployment.
Arkindex: our processing platform
TEKLIA relies on Arkindex, our open source document processing platform, to industrialize large-scale AI projects.
The Arkindex platform makes it possible to structure and automate the processing of heritage collections by integrating a wide range of AI models, manual validation and matching with existing data.
✧ Designed to work on millions of pages
✧ Supports all document types (images, PDF, IIIF)
✧ Articulates processing chains with independent models
✧ Ensures processing traceability and reproducibility
We can integrate emerging AI algorithms and models into Arkindex, ensuring our clients have permanent access to the latest advances and enabling them to continuously improve document processing performance.
Your processing project on Arkindex
For over 10 years, TEKLIA has been working with key players in cultural heritage on ambitious projects involving the automated processing of heritage corpora. Our customized AI solutions automate and improve the analysis, indexing, structuring, and searching of textual and visual cultural corpora, ranging from thousands to millions of documents.
Centralize Your Sources
We collect your documents, regardless of their format or location.
Arkindex organizes your corpora into coherent projects, ready for industrial processing
➙ Flexible bulk import of PDFs and images (JP2, TIFF, JPEG) from your servers, hard drives, or via the IIIF protocol.
The Louvre's documentation in Arkindex.
Reveal Content Through AI
Our algorithms read everything: printed text, ancient manuscripts, and even page structure.
Arkindex integrates generic models or trains custom models (fine-tuning).
➙ Application of high-performance deep learning models for layout segmentation, HTR, and OCR.
Detection of text and illustrations in the excavation notebools of Bernard Bruyère -> IFAO
Enrich the Data
The software does more than read, it understands. It identifies names, places, and dates to classify your documents.
Arkindex interprets document content to create metadata.
➙Transformation of raw text into structured data using Named Entity Recognition (NER) and automatic classification.
Information extraction in an index card in the SIMARA project -> French National Archives
Enhance and Share
Your collections become searchable via a search engine or ready for online publication.
Arkindex enables long-term archiving, integration with your databases, or publication on your web portals.
➙ Standardized export (XML-ALTO, METS, TEI, JSON, CSV) or direct integration via our open API.
Search results by content in the HikarIA platform -> HikarIA
Artificial intelligence technologies for archives, museums and heritage documents
TEKLIA provides a full range of artificial intelligence technologies for document recognition, optimized to extract information from historical and cultural sources.
TEKLIA doesn't develop a one-size-fits-all model. We target each processing task by selecting the most suitable model on the market and implementing it in our workflow. We then integrate these models into Arkindex to build a reliable processing chain tailored to your data.Our processing combines OCR, HTR, computer vision, deep learning models, embedded search, and generative AI.
-> Learn more about our methodology and our commitments
Automatic text recognition (OCR/HTR/ATR)
Convert scanned documents into fully searchable and editable text.
Page layout analysis
Segment the pages into logical areas while preserving the visual hierarchy and context.
Automatic table and data recognition
Extract and convert handwritten or typed tables into usable formats.
Named entities recognition (NER)
Detect and extract names, dates, organizations, places and other important entities from your documents.
Automatic cataloging of collections
Automatically label, classify and index media for integration into digital archives.
Semantic and visual search
Discovering images through their visual attributes or meaning rather than through keywords.
Identify the right technology for your project
Do you want to identify the most suitable processing methods for your documents?Our teams will assist you in choosing AI models and industrializing your project.