CoNLL 2017 and 2018 shared tasks:
Multilingual Parsing from Raw Text to Universal Dependencies
This package contains the test data in the form in which they ware presented
to the participating systems: raw text files and files preprocessed by UDPipe.
The metadata.json files contain lists of files to process and to output;
README files in the respective folders describe the syntax of metadata.json.
For full training, development and gold standard test data, see
Universal Dependencies 2.0 (CoNLL 2017)
Universal Dependencies 2.2 (CoNLL 2018)
See the download links at http://universaldependencies.org/.
For more information on the shared tasks, see
http://universaldependencies.org/conll17/
http://universaldependencies.org/conll18/
Contents:
conll17-ud-test-2017-05-09 ... CoNLL 2017 test data
conll18-ud-test-2018-05-06 ... CoNLL 2018 test data
conll18-ud-test-2018-05-06-for-conll17 ... CoNLL 2018 test data with metadata
and filenames modified so that it is digestible by the 2017 systems.
Baseline UDPipe models for CoNLL 2017 Shared Task in UD Parsing, and supplementary material.
The models require UDPipe version at least 1.1 and are evaluated using the official evaluation script.
The models are trained on a slightly different split of the official UD 2.0 CoNLL 2017 training data, so called baselinemodel split, in order to allow comparison of models even during the shared task. This baselinemodel split of UD 2.0 CoNLL 2017 training data is available for download.
Furthermore, we also provide UD 2.0 CoNLL 2017 training data with automatically predicted morphology. We utilize the baseline models on development data and perform 10-fold jack-knifing (each fold is predicted with a model trained on the rest of the folds) on the training data.
Finally, we supply all required data and hyperparameter values needed to replicate the baseline models.
Baseline UDPipe models for CoNLL 2018 Shared Task in UD Parsing, and supplementary material.
The models require UDPipe version at least 1.2 and are evaluated using the official evaluation script. The models were trained using a custom data split for treebanks where no development data is provided. Also, we trained an additional "Mixed" model, which uses 200 sentences from every training data. All information needed to replicate the model training (hyperparameters, modified train-dev split, and pre-computed word embeddings for the parser) are included in the archive.
Additionaly, we provide UD 2.2 CoNLL 2018 training data with automatically predicted morphology. We utilize the baseline models on development data and perform 10-fold jack-knifing (each fold is predicted with a model trained on the rest of the folds) on the training data.
Czech OOV Inflection Dataset is a Czech inflection dataset of nouns, focused on evaluation in out-of-vocabulary (OOV) conditions. It consists of two parts: a standard lemma-disjoint train-dev-test split of a subset of noun paradigms of existing morphological dictionary Czech MorfFlex 2.0 (files train, dev and test-MorfFlex); and small set of neologisms from Čeština 2.0, annotated for inflected forms (file test-neologisms).
DZ Interset is a means of converting among various tag sets in natural language processing. The core idea is similar to interlingua-based machine translation. DZ Interset defines a set of features that are encoded by the various tag sets. The set of features should be as universal as possible. It does not need to encode everything that is encoded by any tag set but it should encode all information that people may want to access and/or port from one tag set to another.
New tag sets are attached by writing a driver for them. Once the driver is ready, you can easily convert tags between the new set and any other set for which you also have a driver. This reusability is an obvious advantage over writing a targeted conversion procedure each time you need to convert between a particular pair of tag sets. and grant MSM 0021620838 of the Ministry of Education of the Czech Republic
The Feature-based (exponential model) Tagger is a fast implementation of the Czech tagger developed at UFAL and described in the PDT 1.0 documentation (Czech Language Tagging page). In order to get the best possible results, the tagger requires preprocessing by a Czech morphological module with a very high coverage. This module covers a superset of the Czech "FM" morphology. Both the morphological module and the tagger are supplied as binary executables, together with all necessary precompiled Czech data. Input must be in the ISO Latin 2 (iso-8859-2) code and follow the csts.dtd definition, and output is produced in the same way (ISO Latin 2 code, csts.dtd). (As is the case with many of the tools provided with PDT 1.0, both executables also accept - and then produce - a "simplified SGML", which is not a real, valid SGML, but simply contains at least the tags for words, punctuation, and sentence breaks, one item per line.)
The HMM-based Tagger is a software for morphological disambiguation (tagging) of Czech texts. The algorithm is statistical, based on the Hidden Markov Models.
The MORČE tagger is a software for morphological disambiguation (part-of-speech tagging) of Czech text. The algorithm is statistical, based on an idea of so-called "Averaged Perceptron" published by Michael Collins in 2002.
Czech morphological dictionary developed originally by Jan Hajič as a spelling checker and lemmatization dictionary. Currently it contains full morphological information for each covered wordform, as well as some derivational, semantic and named entity information.