The Stanford NLP Group's official Python NLP library. It contains packages for running our latest fully neural pipeline from the CoNLL 2018 Shared Task and for accessing the Java Stanford CoreNLP server. For detailed information please visit our official website.

If you use our neural pipeline including the tokenizer, the multi-word token expansion model, the lemmatizer, the POS/morphological features tagger, or the dependency parser in your research, please kindly cite our CoNLL 2018 Shared Task system description paper:

@inproceedings{qi2018universal,
 address = {Brussels, Belgium},
 author = {Qi, Peng  and  Dozat, Timothy  and  Zhang, Yuhao  and  Manning, Christopher D.},
 booktitle = {Proceedings of the {CoNLL} 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies},
 month = {October},
 pages = {160--170},
 publisher = {Association for Computational Linguistics},
 title = {Universal Dependency Parsing from Scratch},
 url = {https://nlp.stanford.edu/pubs/qi2018universal.pdf},
 year = {2018}
}

The PyTorch implementation of the neural pipeline in this repository is due to Peng Qi and Yuhao Zhang, with help from Tim Dozat and Jason Bolton.

This release is not the same as Stanford's CoNLL 2018 Shared Task system. The tokenizer, lemmatizer, morphological features, and multi-word term systems are a cleaned up version of the shared task code, but in the competition we used a Tensorflow version of the tagger and parser by Tim Dozat, which has been approximately reproduced in PyTorch (though with a few deviations from the original) for this release.

If you use the CoreNLP server, please cite the CoreNLP software package and the respective modules as described here ("Citing Stanford CoreNLP in papers"). The CoreNLP client is mostly written by Arun Chaganty, and Jason Bolton spearheaded merging the two projects together.

To ask questions, report issues or request features, please use the GitHub Issue Tracker.

StanfordNLP supports Python 3.6 or later. We strongly recommend that you install StanfordNLP from PyPI. If you already have pip installed, simply run:

pip install stanfordnlp

this should also help resolve all of the dependencies of StanfordNLP, for instance PyTorch 1.0.0 or above.

If you currently have a previous version of stanfordnlp installed, use:

pip install stanfordnlp -U

Alternatively, you can also install from source of this git repository, which will give you more flexibility in developing on top of StanfordNLP and training your own models. For this option, run

git clone https://github.com/stanfordnlp/stanfordnlp.git
cd stanfordnlp
pip install -e .

To run your first StanfordNLP pipeline, simply following these steps in your Python interactive interpreter:

>>> import stanfordnlp
>>> stanfordnlp.download('en')   # This downloads the English models for the neural pipeline
# IMPORTANT: The above line prompts you before downloading, which doesn't work well in a Jupyter notebook.
# To avoid a prompt when using notebooks, instead use: >>> stanfordnlp.download('en', force=True)
>>> nlp = stanfordnlp.Pipeline() # This sets up a default neural pipeline in English
>>> doc = nlp("Barack Obama was born in Hawaii.  He was elected president in 2008.")
>>> doc.sentences[0].print_dependencies()

The last command will print out the words in the first sentence in the input string (or Document, as it is represented in StanfordNLP), as well as the indices for the word that governs it in the Universal Dependencies parse of that sentence (its "head"), along with the dependency relation between the words. The output should look like:

('Barack', '4', 'nsubj:pass')
('Obama', '1', 'flat')
('was', '4', 'aux:pass')
('born', '0', 'root')
('in', '6', 'case')
('Hawaii', '4', 'obl')
('.', '4', 'punct')

Note: If you are running into issues like OSError: [Errno 22] Invalid argument, it's very likely that you are affected by a known Python issue, and we would recommend Python 3.6.8 or later and Python 3.7.2 or later.

We also provide a multilingual demo script that demonstrates how one uses StanfordNLP in other languages than English, for example Chinese (traditional)

python demo/pipeline_demo.py -l zh

See our getting started guide for more details.

Aside from the neural pipeline, this project also includes an official wrapper for acessing the Java Stanford CoreNLP Server with Python code.

There are a few initial setup steps.

• Download Stanford CoreNLP and models for the language you wish to use
• Put the model jars in the distribution folder
• Tell the python code where Stanford CoreNLP is located: export CORENLP_HOME=/path/to/stanford-corenlp-full-2018-10-05

We provide another demo script that shows how one can use the CoreNLP client and extract various annotations from it.

To get your started, we also provide interactive Jupyter notebooks in the demo folder. You can also open these notebooks and run them interactively on Google Colab. To view all available notebooks, follow these steps:

• Go to the Google Colab website
• Navigate to File -> Open notebook, and choose GitHub in the pop-up menu
• Note that you do not need to give Colab access permission to your github account
• Type stanfordnlp/stanfordnlp in the search bar, and click enter

We currently provide models for all of the treebanks in the CoNLL 2018 Shared Task. You can find instructions for downloading and using these models here.

To maximize speed performance, it is essential to run the pipeline on batches of documents. Running a for loop on one sentence at a time will be very slow. The best approach at this time is to concatenate documents together, with each document separated by a blank line (i.e., two line breaks \n\n). The tokenizer will recognize blank lines as sentence breaks. We are actively working on improving multi-document processing.

All neural modules in this library, including the tokenizer, the multi-word token (MWT) expander, the POS/morphological features tagger, the lemmatizer and the dependency parser, can be trained with your own CoNLL-U format data. Currently, we do not support model training via the Pipeline interface. Therefore, to train your own models, you need to clone this git repository and set up from source.

For detailed step-by-step guidance on how to train and evaluate your own models, please visit our training documentation.

StanfordNLP is released under the Apache License, Version 2.0. See the LICENSE file for more details.