-
Notifications
You must be signed in to change notification settings - Fork 0
Architecture
An installation of Annif may contain multiple independent projects, each of which specifies a set of settings (e.g. language and indexing vocabulary). Projects are limited to a single language; multilingual indexing can be performed by defining multiple projects, one per language.
Each project defines a (typically large) number of subjects, which reflect concepts from an indexing vocabulary. Subjects are typically created from a corpus extracted from existing metadata records and/or indexed documents.
Further, there can be several independent backends that provide analysis functionality. Backends can be integrated into Annif itself, or external services queried via APIs. A project can make use of several backends and combine their analysis results.
Annif has a core application (using Flask and Connexion), which provides both a REST API (when run as a web application) and a command line interface. The command line interface supports administrative functionalities such as corpus management, quality evaluation and other auxiliary functions.
- Home
- Getting started
- System requirements
- Optional features and dependencies
- Usage with Docker
- Architecture
- Commands
- Web user interface
- Corpus formats
- Project configuration
- Analyzers
- Achieving good results
- Reusing preprocessed training data
- Running as a WSGI service
- Backends
- Development flow, branches and tags
- Release process
- Creating a new backend