This submission contains trained end-to-end models for the Neural Monkey toolkit for Czech and English, solving four NLP tasks: machine translation, image captioning, sentiment analysis, and summarization.
The models are trained on standard datasets and achieve state-of-the-art or near state-of-the-art performance in the tasks.
The models are described in the accompanying paper.
The same models can also be invoked via the online demo: https://ufal.mff.cuni.cz/grants/lsd
In addition to the models presented in the referenced paper (developed and published in 2018), we include models for automatic news summarization for Czech and English developed in 2019. The Czech models were trained using the SumeCzech dataset (https://www.aclweb.org/anthology/L18-1551.pdf), the English models were trained using the CNN-Daily Mail corpus (https://arxiv.org/pdf/1704.04368.pdf) using the standard recurrent sequence-to-sequence architecture.
There are several separate ZIP archives here, each containing one model solving one of the tasks for one language.
To use a model, you first need to install Neural Monkey: https://github.com/ufal/neuralmonkey
To ensure correct functioning of the model, please use the exact version of Neural Monkey specified by the commit hash stored in the 'git_commit' file in the model directory.
Each model directory contains a 'run.ini' Neural Monkey configuration file, to be used to run the model. See the Neural Monkey documentation to learn how to do that (you may need to update some paths to correspond to your filesystem organization).
The 'experiment.ini' file, which was used to train the model, is also included.
Then there are files containing the model itself, files containing the input and output vocabularies, etc.
For the sentiment analyzers, you should tokenize your input data using the Moses tokenizer: https://pypi.org/project/mosestokenizer/
For the machine translation, you do not need to tokenize the data, as this is done by the model.
For image captioning, you need to:
- download a trained ResNet: http://download.tensorflow.org/models/resnet_v2_50_2017_04_14.tar.gz
- clone the git repository with TensorFlow models: https://github.com/tensorflow/models
- preprocess the input images with the Neural Monkey 'scripts/imagenet_features.py' script (https://github.com/ufal/neuralmonkey/blob/master/scripts/imagenet_features.py) -- you need to specify the path to ResNet and to the TensorFlow models to this script
The summarization models require input that is tokenized with Moses Tokenizer (https://github.com/alvations/sacremoses) and lower-cased.
Feel free to contact the authors of this submission in case you run into problems!
CzEng 1.0 is the fourth release of a sentence-parallel Czech-English corpus compiled at the Institute of Formal and Applied Linguistics (ÚFAL) freely available for non-commercial research purposes.
CzEng 1.0 contains 15 million parallel sentences (233 million English and 206 million Czech tokens) from seven different types of sources automatically annotated at surface and deep (a- and t-) layers of syntactic representation. and EuroMatrix Plus (FP7-ICT-2007-3-231720 of the EU and 7E09003+7E11051 of the Ministry of Education, Youth and Sports of the Czech Republic),
Faust (FP7-ICT-2009-4-247762 of the EU and 7E11041 of the Ministry of Education, Youth and Sports of the Czech Republic),
GAČR P406/10/P259,
GAUK 116310,
GAUK 4226/2011
This corpora is part of Deliverable 5.5 of the European Commission project QTLeap FP7-ICT-2013.4.1-610516 (http://qtleap.eu).
The texts are sentences from the Europarl parallel corpus (Koehn, 2005). We selected the monolingual sentences from parallel corpora for the following pairs: Bulgarian-English, Czech-English, Portuguese-English and Spanish-English. The English corpus is comprised by the English side of the Spanish-English corpus.
Basque is not in Europarl. In addition, it contains the Basque and English sides of the GNOME corpus.
The texts have been automatically annotated with NLP tools, including Word Sense Disambiguation, Named Entity Disambiguation and Coreference resolution. Please check deliverable D5.6 in http://qtleap.eu/deliverables for more information.
This machine translation test set contains 2223 Czech sentences collected within the FAUST project (https://ufal.mff.cuni.cz/grants/faust, http://hdl.handle.net/11234/1-3308).
Each original (noisy) sentence was normalized (clean1 and clean2) and translated to English independently by two translators.
HamleDT (HArmonized Multi-LanguagE Dependency Treebank) is a compilation of existing dependency treebanks (or dependency conversions of other treebanks), transformed so that they all conform to the same annotation style. This version uses Universal Dependencies as the common annotation style.
Update (November 1017): for a current collection of harmonized dependency treebanks, we recommend using the Universal Dependencies (UD). All of the corpora that are distributed in HamleDT in full are also part of the UD project; only some corpora from the Patch group (where HamleDT provides only the harmonizing scripts but not the full corpus data) are available in HamleDT but not in UD.
Source code of the LINDAT Translation service frontend. The service provides a UI and a simple rest api that accesses machine translation models served by tensorflow serving.
The most recent version of the code is available at https://github.com/ufal/lindat_translation.
The collection comprises the relevance judgments used in the 2023 LongEval Information Retrieval Lab (https://clef-longeval.github.io/), organized at CLEF. It consists of three sets of relevance judgments:
1) Relevance judgments for the heldout queries from the LongEval Train Collection (http://hdl.handle.net/11234/1-5010).
2) Relevance judgments for the short-term persistence (sub-task A) queries from the LongEval Test Collection (http://hdl.handle.net/11234/1-5139).
3) Relevance judgments for the long-term persistence (sub-task B) queries from the LongEval Test Collection (http://hdl.handle.net/11234/1-5139).
These judgments were provided by the Qwant search engine (https://www.qwant.com) and were generated using a click model. The click model output was based on the clicks of Qwant's users, but it mitigates noise from raw user clicks caused by positional bias and also better safeguards users' privacy. Consequently, it can serve as a reliable soft relevance estimate for evaluating and training models.
The collection includes a total of 1,420 judgments for the heldout queries, with 74 considered highly relevant and 326 deemed relevant. For the short-term sub-task queries, there are 12,217 judgments, including 762 highly relevant and 2,608 relevant ones. As for the long-term sub-task queries, there are 13,467 judgments, with 936 being highly relevant and 2,899 relevant.
The collection consists of queries and documents provided by the Qwant search Engine (https://www.qwant.com). The queries, which were issued by the users of Qwant, are based on the selected trending topics. The documents in the collection are the webpages which were selected with respect to these queries using the Qwant click model. Apart from the documents selected using this model, the collection also contains randomly selected documents from the Qwant index.
The collection serves as the official test collection for the 2023 LongEval Information Retrieval Lab (https://clef-longeval.github.io/) organised at CLEF. The collection contains test datasets for two organized sub-tasks: short-term persistence (sub-task A) and long-term persistence (sub-task B). The data for the short-term persistence sub-task was collected over July 2022 and this dataset contains 1,593,376 documents and 882 queries. The data for the long-term persistence sub-task was collected over September 2022 and this dataset consists of 1,081,334 documents and 923 queries. Apart from the original French versions of the webpages and queries, the collection also contains their translations into English.
The collection consists of queries and documents provided by the Qwant search Engine (https://www.qwant.com). The queries, which were issued by the users of Qwant, are based on the selected trending topics. The documents in the collection were selected with respect to these queries using the Qwant click model. Apart from the documents selected using this model, the collection also contains randomly selected documents from the Qwant index. All the data were collected over June 2022. In total, the collection contains 672 train queries, with corresponding 9656 assessments coming from the Qwant click model, and 98 heldout queries. The set of documents consist of 1,570,734 downloaded, cleaned and filtered Web Pages. Apart from their original French versions, the collection also contains translations of the webpages and queries into English. The collection serves as the official training collection for the 2023 LongEval Information Retrieval Lab (https://clef-longeval.github.io/) organised at CLEF.
This data set contains four types of manual annotation of translation quality, focusing on the comparison of human and machine translation quality (aka human-parity). The machine translation system used is English-Czech CUNI Transformer (CUBBITT). The annotations distinguish adequacy, fluency and overall quality. One of the types is Translation Turing test - detecting whether the annotators can distinguish human from machine translation.
All the sentences are taken from the English-Czech test set newstest2018 (WMT2018 News translation shared task www.statmt.org/wmt18/translation-task.html), but only from the half with originally English sentences translated to Czech by a professional agency.