flexiPipe is a modular NLP pipeline for Universal Dependencies data: rule-based components, the flexitag tagger, and multiple neural backends (SpaCy, Stanza, Flair, UDPipe, UDMorph, UD-Kanbun, etc.). It ingests raw text, CoNLL-U, and TEITOK XML and exports CoNLL-U, TEITOK, SVG/HTML/LaTeX trees, and more.
Full documentation: see the wiki/ folder — wiki/README.md is the index; installation, backends, quick start, and reference are in the linked pages.
pip install flexipipe
# Optional backends (use flexipipe info backends to see all):
flexipipe install spacy
flexipipe install udapi # for HTML/LaTeX outputThen run flexipipe config --wizard to pick defaults. To run as flexipipe instead of python -m flexipipe, install the wrapper: python -m flexipipe install wrapper (see Installation).
# Tag text (default backend)
echo "Don't even think he wouldn't do it." | flexipipe process --input -
# With a backend and model
flexipipe process --backend spacy --model en_core_web_sm --input file.txt --output out.conllu
# See available backends and subcommand help
flexipipe info backends
flexipipe process --help| Command | Purpose |
|---|---|
flexipipe process |
Tag/parse/normalize input |
flexipipe benchmark |
Evaluate backends on gold data |
flexipipe train |
Train flexitag or backend models |
flexipipe convert |
Convert formats (e.g. TEITOK → CoNLL-U) |
flexipipe info |
List backends, models, examples, tasks, languages, installation |
flexipipe config |
Set defaults (backend, output format, models dir) |
flexipipe install |
Install optional backends or the launcher; upgrade flexipipe |
Use flexipipe <subcommand> --help for full options. Use flexipipe info backends to see available backends (more can be installed via flexipipe install <name>).
flexipipe/— main package (CLI, backends, converters)flexitag/— C++ flexitag sources and bindingsREADME_CPP.md— native build instructionswiki/— full documentation (Markdown); wiki/README.md is the index.
See Contributing for development notes.