Automatic indexing of your documents, in France
Automatically assign subject keywords to your documents with Annif, installed and maintained by DINAO. Your corpora and indexing models stay on our French servers — never sent to a third-party AI service.
What is Annif?
Annif is a multi-algorithm automatic subject indexing tool designed for libraries, archives, and museums. Developed by the National Library of Finland, it automatically assigns subject keywords to documents from a controlled vocabulary, significantly speeding up cataloging work.
Annif combines several natural language processing and machine learning tools — TensorFlow, Omikuji, fastText, spaCy — through various backends (TF-IDF, fastText, MLLM, neural network ensembles) that can be mixed to achieve the best results. It is multilingual and supports any subject vocabulary in SKOS or TSV format.
The tool provides a command-line interface, a simple web interface, and a microservice-style REST API, facilitating integration into existing documentary systems. Published under the Apache 2.0 license, Annif runs its models locally: your corpora and models remain fully under your control.
Host Annif at DINAO
Resource tiers compatible with Annif prerequisites (minimum 1 vCPU / 2 Go / 5 Go). Hosted in France, fully managed.
- 1 dedicated vCPU
- 2 Go RAM
- 20 GB NVMe
- Daily backups
- Managed & monitored by DINAO
- 2 dedicated vCPU
- 4 Go RAM
- 40 GB NVMe
- Daily backups
- Managed & monitored by DINAO
- 4 dedicated vCPU
- 8 Go RAM
- 80 GB NVMe
- Daily backups
- Managed & monitored by DINAO
- 8 dedicated vCPU
- 16 Go RAM
- 160 GB NVMe
- Daily backups
- Managed & monitored by DINAO
Technical details
You might be wondering…
What is Annif used for?
Annif automatically assigns subject keywords (topics) to your documents from a controlled vocabulary. It is designed for libraries, archives, and museums that want to quickly index large collections.
Does Annif send my documents to an AI service?
No. Annif runs entirely on your DINAO instance in France. Models are trained and executed locally from your own corpora — no corpora or documents are sent to a third-party AI service.
Where is the data hosted?
On DINAO infrastructure in France, in one of the available datacenters. Your training corpora, vocabularies, and models never leave the territory.
Are technical skills required?
For everyday use (indexing via the API or web interface), no. However, preparing vocabularies and training models require a documentary or data profile; DINAO manages hosting, dependencies, and updates.
Does Annif work in French and with my vocabulary?
Yes. Annif is multilingual and accepts any subject vocabulary in SKOS or TSV format, allowing it to be adapted to your thesaurus and working languages.
