Skip to content
Snippets Groups Projects

Installation

Currently, COMBO is available as a git repository. In the cloned repository directory:

(Recommended) Create a Conda environment

conda create -n combo python=3.9.16

Make sure to install at least python 3.9.16.

Not all of the requirements are available on conda, so pip is also required.

conda install pip

Automatic installation

pip install combo-nlp

Source installation

Install the requirements using the setup.py file.

pip install -e .

Use COMBO as a python package

import combo

Pretrained model usage:

from combo.predict import COMBO
c = COMBO.from_pretrained('polish-herbert-base-ud213')
prediction = c("To jest przykładowe zdanie. To jest drugie zdanie.")

# Display the first sentence predictions
print("{:15} {:15} {:10} {:10} {:10}".format('TOKEN', 'LEMMA', 'UPOS', 'HEAD', 'DEPREL'))
for token in prediction[0].tokens:
    print("{:15} {:15} {:10} {:10} {:10}".format(token.text, token.lemma, token.upostag, token.head, token.deprel))

Example output:

TOKEN           LEMMA           UPOS       HEAD       DEPREL
To              to              AUX                 4 cop       
jest            być             AUX                 4 cop       
przykładowe     przykładowy     ADJ                 4 amod      
zdanie          zdanie          NOUN                0 root      
.               .               PUNCT               4 punct  

Use COMBO CLI

The minimal training example (make sure to download some conllu training and validation files)

combo --mode train --training_data_path <training conllu> --validation_data_path <validation conllu>

You can find more examples in docs/Training.md

Perform Named Entity Recognition with COMBO

COMBO has a NER module. Currently, three languages are supported: English (base and large version), Polish (base and large version) and Spanish (base version). However, it is possible to train your own NER models. Refer to documentation for more information here.

Performance of the pretrained models:

name dataset language F1 score
on devset
F1 score
on testset
Precision
on devset
Precision
on testset
Recall
on devset
Recall
on testset
pl_base kpwe-n82 polish 73,58 72,39 73,58 72,84 73,58 71,94
pl_large kpwe-n82 polish 74,97 74,34 74,70 74,59 75,24 74,089
eng_base ConLL03 english 95,25 92,06 95,20 92,42 95,30 91,69
eng_large ConLL03 english 95,17 92,11 95,04 92,49 95,30 91,73
es_base ConLL02 spanish 85,48 87,38 85,89 87,01 85,08 87,74

More granular data about performance can be found here

There are two example notebooks that show how to use the NER module:

  • Training a NER model here
  • Using a pretrained NER model here

COMBO tutorial

We encourage you to use the beginner's tutorial (colab notebook).

Details

Citing

If you use COMBO in your research, please cite COMBO: State-of-the-Art Morphosyntactic Analysis

@inproceedings{klimaszewski-wroblewska-2021-combo-state,
    title = "{COMBO}: State-of-the-Art Morphosyntactic Analysis",
    author = "Klimaszewski, Mateusz  and
      Wr{\'o}blewska, Alina",
    booktitle = "Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing: System Demonstrations",
    month = nov,
    year = "2021",
    address = "Online and Punta Cana, Dominican Republic",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.emnlp-demo.7",
    pages = "50--62",
    abstract = "We introduce COMBO {--} a fully neural NLP system for accurate part-of-speech tagging, morphological analysis, lemmatisation, and (enhanced) dependency parsing. It predicts categorical morphosyntactic features whilst also exposes their vector representations, extracted from hidden layers. COMBO is an easy to install Python package with automatically downloadable pre-trained models for over 40 languages. It maintains a balance between efficiency and quality. As it is an end-to-end system and its modules are jointly trained, its training is competitively fast. As its models are optimised for accuracy, they achieve often better prediction quality than SOTA. The COMBO library is available at: https://gitlab.clarin-pl.eu/syntactic-tools/combo.",
}

If you use an EUD module in your research, please cite COMBO: A New Module for EUD Parsing

@inproceedings{klimaszewski-wroblewska-2021-combo,
    title = "{COMBO}: A New Module for {EUD} Parsing",
    author = "Klimaszewski, Mateusz  and
      Wr{\'o}blewska, Alina",
    booktitle = "Proceedings of the 17th International Conference on Parsing Technologies and the IWPT 2021 Shared Task on Parsing into Enhanced Universal Dependencies (IWPT 2021)",
    month = aug,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.iwpt-1.16",
    doi = "10.18653/v1/2021.iwpt-1.16",
    pages = "158--166",
    abstract = "We introduce the COMBO-based approach for EUD parsing and its implementation, which took part in the IWPT 2021 EUD shared task. The goal of this task is to parse raw texts in 17 languages into Enhanced Universal Dependencies (EUD). The proposed approach uses COMBO to predict UD trees and EUD graphs. These structures are then merged into the final EUD graphs. Some EUD edge labels are extended with case information using a single language-independent expansion rule. In the official evaluation, the solution ranked fourth, achieving an average ELAS of 83.79{\%}. The source code is available at https://gitlab.clarin-pl.eu/syntactic-tools/combo.",
}