Udpipe

UDPipe is an open-source natural language processing toolkit designed to process text in alignment with the Universal Dependencies framework. It provides a trainable pipeline that can tokenize, segment sentences, perform part-of-speech tagging, lemmatization, morphological analysis, and dependency parsing. The resulting annotations are compatible with the CoNLL-U format used by UD treebanks, enabling standardized linguistic annotation across languages.

The tool combines language-specific models with a general architecture that supports training on user-provided data. Users can apply pre-trained models to annotate raw text or train new models on UD treebanks to create language-specific parsers. UDPipe is implemented in C++ and offers bindings for Python and Java, making it usable as a command-line utility or as a library within larger NLP workflows.

UDPipe provides models for many languages covered by the Universal Dependencies project, allowing researchers and developers to annotate text out of the box for a wide range of languages. It also supports training custom models from UD data, facilitating experiments in language-specific morphology and syntax. The output can be used directly for downstream tasks such as parsing, corpus annotation, or linguistic research, and can be converted or integrated into other UD processing pipelines.

Availability and licensing are governed by its open-source status, with distribution through the official UDPipe resources and the Universal Dependencies project. The toolkit is commonly used in academic and research contexts for linguistic annotation, treebank construction, and evaluation of parsing and tagging models.