Milan Straka

Milan Straka

We introduce CorPipe 26, our winning submission to the CRAC 2026 Shared Task on Multilingual Coreference Resolution. The fifth edition of this shared task focuses mainly on the comparison of generative LLMs and specialized systems; additionally, 5 more datasets and 2 new languages are introduced. CorPipe 26 is an improved version of CorPipe 25, with a new variant predicting empty nodes together with mentions and coreference links in a single model. Our system outperforms all other submissions in the LLM track by 2.8 percent points and all submissions in the unconstrained track by 9.5 percent points. Furthermore, we perform a series of ablation experiments with different model sizes, empty node prediction methods, and cross-lingual zero-shot evaluation. The source code and the trained models are publicly available at https://github.com/ufal/crac2026-corpipe.

Milan Straka, Jana Straková

We describe the winning submission to the CRAC 2022 Shared Task on Multilingual Coreference Resolution. Our system first solves mention detection and then coreference linking on the retrieved spans with an antecedent-maximization approach, and both tasks are fine-tuned jointly with shared Transformer weights. We report results of fine-tuning a wide range of pretrained models. The center of this contribution are fine-tuned multilingual models. We found one large multilingual model with sufficiently large encoder to increase performance on all datasets across the board, with the benefit not limited only to the underrepresented languages or groups of typologically relative languages. The source code is available at https://github.com/ufal/crac2022-corpipe.

Marie Mikulová, Milan Straka, Jan Štěpánek et al.

This paper presents an analysis of annotation using an automatic pre-annotation for a mid-level annotation complexity task -- dependency syntax annotation. It compares the annotation efforts made by annotators using a pre-annotated version (with a high-accuracy parser) and those made by fully manual annotation. The aim of the experiment is to judge the final annotation quality when pre-annotation is used. In addition, it evaluates the effect of automatic linguistically-based (rule-formulated) checks and another annotation on the same data available to the annotators, and their influence on annotation quality and efficiency. The experiment confirmed that the pre-annotation is an efficient tool for faster manual syntactic annotation which increases the consistency of the resulting annotation without reducing its quality.

Milan Straka

We present CorPipe, the winning entry to the CRAC 2023 Shared Task on Multilingual Coreference Resolution. Our system is an improved version of our earlier multilingual coreference pipeline, and it surpasses other participants by a large margin of 4.5 percent points. CorPipe first performs mention detection, followed by coreference linking via an antecedent-maximization approach on the retrieved spans. Both tasks are trained jointly on all available corpora using a shared pretrained language model. Our main improvements comprise inputs larger than 512 subwords and changing the mention decoding to support ensembling. The source code is available at https://github.com/ufal/crac2023-corpipe.

David Kubeša, Milan Straka

We present DaMuEL, a large Multilingual Dataset for Entity Linking containing data in 53 languages. DaMuEL consists of two components: a knowledge base that contains language-agnostic information about entities, including their claims from Wikidata and named entity types (PER, ORG, LOC, EVENT, BRAND, WORK_OF_ART, MANUFACTURED); and Wikipedia texts with entity mentions linked to the knowledge base, along with language-specific text from Wikidata such as labels, aliases, and descriptions, stored separately for each language. The Wikidata QID is used as a persistent, language-agnostic identifier, enabling the combination of the knowledge base with language-specific texts and information for each entity. Wikipedia documents deliberately annotate only a single mention for every entity present; we further automatically detect all mentions of named entities linked from each document. The dataset contains 27.9M named entities in the knowledge base and 12.3G tokens from Wikipedia texts. The dataset is published under the CC BY-SA license at https://hdl.handle.net/11234/1-5047.

Michal Novák, Miloslav Konopík, Anna Nedoluzhko et al.

This paper describes the fifth edition of the Shared Task on Multilingual Coreference Resolution, held in conjunction with the CODI-CRAC 2026 workshop. Building on previous iterations, the task required participants to develop systems capable of mention identification and identity-based coreference clustering. The 2026 edition specifically emphasizes long-range entities, defined as coreferential chains spanning significant distances, across many words and sentences. The task expanded its linguistic scope by incorporating five new datasets and two additional languages. These additions leverage version 1.4 of CorefUD, a harmonized multilingual collection comprising 27 datasets in 19 languages. In total, ten systems participated, including four LLM-based approaches (three fine-tuned models and one few-shot approach). While traditional systems still maintained their lead, LLMs demonstrated significant potential, suggesting they may soon challenge established approaches in future editions.

Milan Straka, Jana Straková, Federica Gamba

We present LatinPipe, the winning submission to the EvaLatin 2024 Dependency Parsing shared task. Our system consists of a fine-tuned concatenation of base and large pre-trained LMs, with a dot-product attention head for parsing and softmax classification heads for morphology to jointly learn both dependency parsing and morphological analysis. It is trained by sampling from seven publicly available Latin corpora, utilizing additional harmonization of annotations to achieve a more unified annotation style. Before fine-tuning, we train the system for a few initial epochs with frozen weights. We also add additional local relative contextualization by stacking the BiLSTM layers on top of the Transformer(s). Finally, we ensemble output probability distributions from seven randomly instantiated networks for the final submission. The code is available at https://github.com/ufal/evalatin2024-latinpipe.

Milan Straka

We present CorPipe 25, the winning entry to the CRAC 2025 Shared Task on Multilingual Coreference Resolution. This fourth iteration of the shared task introduces a new LLM track alongside the original unconstrained track, features reduced development and test sets to lower computational requirements, and includes additional datasets. CorPipe 25 represents a complete reimplementation of our previous systems, migrating from TensorFlow to PyTorch. Our system significantly outperforms all other submissions in both the LLM and unconstrained tracks by a substantial margin of 8 percentage points. The source code and trained models are publicly available at https://github.com/ufal/crac2025-corpipe.

Petr Pechman, Milan Straka, Jana Straková et al.

We present a grammar error correction (GEC) system that achieves state of the art for the Czech language. Our system is based on a neural network translation approach with the Transformer architecture, and its key feature is its real-time synthetic generation pipeline, which dynamically augments sentences with artificial errors by introducing both language-agnostic and Czech-specific errors. We conduct a comprehensive series of experiments, investigating the Czech GEC corpora as bases for synthetic error introduction, several error generation strategies, domain balancing, tokenization granularity, model size, and data scaling during fine-tuning. Additionally, we evaluate the performance of large language models (LLMs) on Czech GEC in both end-user and expert fine-tuning scenarios. Our best-performing model is superior both in performance and computational efficiency. The source code and the trained model links are available on https://github.com/ufal/tsd2025-gec.

Milan Straka, Jana Straková

We present an open-source web service for Czech morphosyntactic analysis. The system combines a deep learning model with rescoring by a high-precision morphological dictionary at inference time. We show that our hybrid method surpasses two competitive baselines: While the deep learning model ensures generalization for out-of-vocabulary words and better disambiguation, an improvement over an existing morphological analyser MorphoDiTa, at the same time, the deep learning model benefits from inference-time guidance of a manually curated morphological dictionary. We achieve 50% error reduction in lemmatization and 58% error reduction in POS tagging over MorphoDiTa, while also offering dependency parsing. The model is trained on one of the currently largest Czech morphosyntactic corpora, the PDT-C 1.0, with the trained models available at https://hdl.handle.net/11234/1-5293. We provide the tool as a web service deployed at https://lindat.mff.cuni.cz/services/udpipe/. The source code is available at GitHub (https://github.com/ufal/udpipe/tree/udpipe-2), along with a Python client for a simple use. The documentation for the models can be found at https://ufal.mff.cuni.cz/udpipe/2/models#czech_pdtc1.0_model.

Milan Straka, Jakub Náplava, Jana Straková

We propose a character-based nonautoregressive GEC approach, with automatically generated character transformations. Recently, per-word classification of correction edits has proven an efficient, parallelizable alternative to current encoder-decoder GEC systems. We show that word replacement edits may be suboptimal and lead to explosion of rules for spelling, diacritization and errors in morphologically rich languages, and propose a method for generating character transformations from GEC corpus. Finally, we train character transformation models for Czech, German and Russian, reaching solid results and dramatic speedup compared to autoregressive systems. The source code is released at https://github.com/ufal/wnut2021_character_transformations_gec.

David Samuel, Milan Straka

We present the winning entry to the Multilingual Lexical Normalization (MultiLexNorm) shared task at W-NUT 2021 (van der Goot et al., 2021a), which evaluates lexical-normalization systems on 12 social media datasets in 11 languages. We base our solution on a pre-trained byte-level language model, ByT5 (Xue et al., 2021a), which we further pre-train on synthetic data and then fine-tune on authentic normalization data. Our system achieves the best performance by a wide margin in intrinsic evaluation, and also the best performance in extrinsic evaluation through dependency parsing. The source code is released at https://github.com/ufal/multilexnorm2021 and the fine-tuned models at https://huggingface.co/ufal.

Jakub Náplava, Martin Popel, Milan Straka et al.

Sensitivity of deep-neural models to input noise is known to be a challenging problem. In NLP, model performance often deteriorates with naturally occurring noise, such as spelling errors. To mitigate this issue, models may leverage artificially noised data. However, the amount and type of generated noise has so far been determined arbitrarily. We therefore propose to model the errors statistically from grammatical-error-correction corpora. We present a thorough evaluation of several state-of-the-art NLP systems' robustness in multiple languages, with tasks including morpho-syntactic analysis, named entity recognition, neural machine translation, a subset of the GLUE benchmark and reading comprehension. We also compare two approaches to address the performance drop: a) training the NLP models with noised data generated by our framework; and b) reducing the input noise with external system for natural language correction. The code is released at https://github.com/ufal/kazitext.

Jakub Náplava, Milan Straka, Jana Straková

We propose a new architecture for diacritics restoration based on contextualized embeddings, namely BERT, and we evaluate it on 12 languages with diacritics. Furthermore, we conduct a detailed error analysis on Czech, a morphologically rich language with a high level of diacritization. Notably, we manually annotate all mispredictions, showing that roughly 44% of them are actually not errors, but either plausible variants (19%), or the system corrections of erroneous data (25%). Finally, we categorize the real errors in detail. We release the code at https://github.com/ufal/bert-diacritics-restoration.

Milan Straka, Jakub Náplava, Jana Straková et al.

We present RobeCzech, a monolingual RoBERTa language representation model trained on Czech data. RoBERTa is a robustly optimized Transformer-based pretraining approach. We show that RobeCzech considerably outperforms equally-sized multilingual and Czech-trained contextualized language representation models, surpasses current state of the art in all five evaluated NLP tasks and reaches state-of-the-art results in four of them. The RobeCzech model is released publicly at https://hdl.handle.net/11234/1-3691 and https://huggingface.co/ufal/robeczech-base.

David Samuel, Milan Straka

We present PERIN, a novel permutation-invariant approach to sentence-to-graph semantic parsing. PERIN is a versatile, cross-framework and language independent architecture for universal modeling of semantic structures. Our system participated in the CoNLL 2020 shared task, Cross-Framework Meaning Representation Parsing (MRP 2020), where it was evaluated on five different frameworks (AMR, DRG, EDS, PTG and UCCA) across four languages. PERIN was one of the winners of the shared task. The source code and pretrained models are available at https://github.com/ufal/perin.

Milan Straka, Jana Straková

We present a system description of our contribution to the CoNLL 2019 shared task, Cross-Framework Meaning Representation Parsing (MRP 2019). The proposed architecture is our first attempt towards a semantic parsing extension of the UDPipe 2.0, a lemmatization, POS tagging and dependency parsing pipeline. For the MRP 2019, which features five formally and linguistically different approaches to meaning representation (DM, PSD, EDS, UCCA and AMR), we propose a uniform, language and framework agnostic graph-to-graph neural network architecture. Without any knowledge about the graph structure, and specifically without any linguistically or framework motivated features, our system implicitly models the meaning representation graphs. After fixing a human error (we used earlier incorrect version of provided test set analyses), our submission would score third in the competition evaluation. The source code of our system is available at https://github.com/ufal/mrpipe-conll2019.

Jakub Náplava, Milan Straka

Grammatical error correction in English is a long studied problem with many existing systems and datasets. However, there has been only a limited research on error correction of other languages. In this paper, we present a new dataset AKCES-GEC on grammatical error correction for Czech. We then make experiments on Czech, German and Russian and show that when utilizing synthetic parallel corpus, Transformer neural machine translation model can reach new state-of-the-art results on these datasets. AKCES-GEC is published under CC BY-NC-SA 4.0 license at https://hdl.handle.net/11234/1-3057 and the source code of the GEC model is available at https://github.com/ufal/low-resource-gec-wnut2019.

Dan Kondratyuk, Milan Straka

We present UDify, a multilingual multi-task model capable of accurately predicting universal part-of-speech, morphological features, lemmas, and dependency trees simultaneously for all 124 Universal Dependencies treebanks across 75 languages. By leveraging a multilingual BERT self-attention model pretrained on 104 languages, we found that fine-tuning it on all datasets concatenated together with simple softmax classifiers for each UD task can result in state-of-the-art UPOS, UFeats, Lemmas, UAS, and LAS scores, without requiring any recurrent or language-specific components. We evaluate UDify for multilingual learning, showing that low-resource languages benefit the most from cross-linguistic annotations. We also evaluate for zero-shot learning, with results suggesting that multilingual training provides strong UD predictions even for languages that neither UDify nor BERT have ever been trained on. Code for UDify is available at https://github.com/hyperparticle/udify.

Jiří Mayer, Milan Straka, Jan Hajič et al.

The majority of recent progress in Optical Music Recognition (OMR) has been achieved with Deep Learning methods, especially models following the end-to-end paradigm, reading input images and producing a linear sequence of tokens. Unfortunately, many music scores, especially piano music, cannot be easily converted to a linear sequence. This has led OMR researchers to use custom linearized encodings, instead of broadly accepted structured formats for music notation. Their diversity makes it difficult to compare the performance of OMR systems directly. To bring recent OMR model progress closer to useful results: (a) We define a sequential format called Linearized MusicXML, allowing to train an end-to-end model directly and maintaining close cohesion and compatibility with the industry-standard MusicXML format. (b) We create a dev and test set for benchmarking typeset OMR with MusicXML ground truth based on the OpenScore Lieder corpus. They contain 1,438 and 1,493 pianoform systems, each with an image from IMSLP. (c) We train and fine-tune an end-to-end model to serve as a baseline on the dataset and employ the TEDn metric to evaluate the model. We also test our model against the recently published synthetic pianoform dataset GrandStaff and surpass the state-of-the-art results.

Michal Novák, Barbora Dohnalová, Miloslav Konopík et al.

The paper presents an overview of the third edition of the shared task on multilingual coreference resolution, held as part of the CRAC 2024 workshop. Similarly to the previous two editions, the participants were challenged to develop systems capable of identifying mentions and clustering them based on identity coreference. This year's edition took another step towards real-world application by not providing participants with gold slots for zero anaphora, increasing the task's complexity and realism. In addition, the shared task was expanded to include a more diverse set of languages, with a particular focus on historical languages. The training and evaluation data were drawn from version 1.2 of the multilingual collection of harmonized coreference resources CorefUD, encompassing 21 datasets across 15 languages. 6 systems competed in this shared task.

Michal Novák, Miloslav Konopík, Anna Nedoluzhko et al.

The paper presents an overview of the fourth edition of the Shared Task on Multilingual Coreference Resolution, organized as part of the CODI-CRAC 2025 workshop. As in the previous editions, participants were challenged to develop systems that identify mentions and cluster them according to identity coreference. A key innovation of this year's task was the introduction of a dedicated Large Language Model (LLM) track, featuring a simplified plaintext format designed to be more suitable for LLMs than the original CoNLL-U representation. The task also expanded its coverage with three new datasets in two additional languages, using version 1.3 of CorefUD - a harmonized multilingual collection of 22 datasets in 17 languages. In total, nine systems participated, including four LLM-based approaches (two fine-tuned and two using few-shot adaptation). While traditional systems still kept the lead, LLMs showed clear potential, suggesting they may soon challenge established approaches in future editions.

Matěj Kripner, Michal Šustr, Milan Straka

Automated theorem proving (ATP) has been a classical problem in artificial intelligence since its inception, yet it remains challenging due to its vast state and action space. Large language models (LLMs) have recently emerged as a promising heuristic for ATP, but they lack correctness guarantees and thus require interaction with a proof verifier. Such interactions typically follow one of two approaches: black-box interaction, which does not utilize intermediate proof states, or white-box approaches, which allow for incremental proof construction and examination of intermediate states. While black-box approaches have directly benefited from recent LLM advances, white-box methods have comparatively lagged behind. In this paper, we address this gap by introducing LeanTree, which consists of (i) a tool built in the Lean 4 language that factorizes complex proof states into simpler, independent branches, and (ii) a dataset of these factorized intermediate states. Our white-box tooling offers several advantages over black-box approaches: it simplifies evaluation, reduces necessary context, generates richer training data, enables parallel search across multiple states, supports efficient reuse of states, and provides feedback in case of errors. Our preliminary results hint that white-box approaches outperform black-box alternatives in some settings.

Jakub Náplava, Milan Straka, Jana Straková et al.

We introduce a large and diverse Czech corpus annotated for grammatical error correction (GEC) with the aim to contribute to the still scarce data resources in this domain for languages other than English. The Grammar Error Correction Corpus for Czech (GECCC) offers a variety of four domains, covering error distributions ranging from high error density essays written by non-native speakers, to website texts, where errors are expected to be much less common. We compare several Czech GEC systems, including several Transformer-based ones, setting a strong baseline to future research. Finally, we meta-evaluate common GEC metrics against human judgements on our data. We make the new Czech GEC corpus publicly available under the CC BY-SA 4.0 license at http://hdl.handle.net/11234/1-4639 .

Milan Straka, Lucia Piatriková, Peter van Bokhoven et al.

Based on the electric vehicle (EV) arrival times and the duration of EV connection to the charging station, we identify charging patterns and derive groups of charging stations with similar charging patterns applying two approaches. The ruled based approach derives the charging patterns by specifying a set of time intervals and a threshold value. In the second approach, we combine the modified l-p norm (as a matrix dissimilarity measure) with hierarchical clustering and apply them to automatically identify charging patterns and groups of charging stations associated with such patterns. A dataset collected in a large network of public charging stations is used to test both approaches. Using both methods, we derived charging patterns. The first, rule-based approach, performed well at deriving predefined patterns and the latter, hierarchical clustering, showed the capability of delivering unexpected charging patterns.

Kateřina Macková, Milan Straka

Reading comprehension is a well studied task, with huge training datasets in English. This work focuses on building reading comprehension systems for Czech, without requiring any manually annotated Czech training data. First of all, we automatically translated SQuAD 1.1 and SQuAD 2.0 datasets to Czech to create training and development data, which we release at http://hdl.handle.net/11234/1-3249. We then trained and evaluated several BERT and XLM-RoBERTa baseline models. However, our main focus lies in cross-lingual transfer models. We report that a XLM-RoBERTa model trained on English data and evaluated on Czech achieves very competitive performance, only approximately 2 percent points worse than a~model trained on the translated Czech data. This result is extremely good, considering the fact that the model has not seen any Czech data during training. The cross-lingual transfer approach is very flexible and provides a reading comprehension in any language, for which we have enough monolingual raw texts.

Milan Straka, Jana Straková

We present our contribution to the EvaLatin shared task, which is the first evaluation campaign devoted to the evaluation of NLP tools for Latin. We submitted a system based on UDPipe 2.0, one of the winners of the CoNLL 2018 Shared Task, The 2018 Shared Task on Extrinsic Parser Evaluation and SIGMORPHON 2019 Shared Task. Our system places first by a wide margin both in lemmatization and POS tagging in the open modality, where additional supervised data is allowed, in which case we utilize all Universal Dependency Latin treebanks. In the closed modality, where only the EvaLatin training data is allowed, our system achieves the best performance in lemmatization and in classical subtask of POS tagging, while reaching second place in cross-genre and cross-time settings. In the ablation experiments, we also evaluate the influence of BERT and XLM-RoBERTa contextualized embeddings, and the treebank encodings of the different flavors of Latin treebanks.

Jan Hajič, Eduard Bejček, Jaroslava Hlaváčová et al.

We present a richly annotated and genre-diversified language resource, the Prague Dependency Treebank-Consolidated 1.0 (PDT-C 1.0), the purpose of which is - as it always been the case for the family of the Prague Dependency Treebanks - to serve both as a training data for various types of NLP tasks as well as for linguistically-oriented research. PDT-C 1.0 contains four different datasets of Czech, uniformly annotated using the standard PDT scheme (albeit not everything is annotated manually, as we describe in detail here). The texts come from different sources: daily newspaper articles, Czech translation of the Wall Street Journal, transcribed dialogs and a small amount of user-generated, short, often non-standard language segments typed into a web translator. Altogether, the treebank contains around 180,000 sentences with their morphological, surface and deep syntactic annotation. The diversity of the texts and annotations should serve well the NLP applications as well as it is an invaluable resource for linguistic research, including comparative studies regarding texts of different genres. The corpus is publicly and freely available.

Milan Straka, Rui Carvalho, Gijs van der Poel et al.

Here, we develop a data-centric approach enabling to analyse which activities, function, and characteristics of the environment surrounding the slow charging infrastructure impact the distribution of the electricity consumed at slow charging infrastructure. To gain a basic insight, we analysed the probabilistic distribution of energy consumption and its relation to indicators characterizing charging events. We collected geospatial datasets and utilizing statistical methods for data pre-processing, we prepared features modelling the spatial context in which the charging infrastructure operates. To enhance the statistical reliability of results, we applied the bootstrap method together with the Lasso method that combines regression with variable selection ability. We evaluate the statistical distributions of the selected regression coefficients. We identified the most influential features correlated with energy consumption, indicating that the spatial context of the charging infrastructure affects its utilization pattern. Many of these features are related to the economic prosperity of residents. Application of the methodology to a specific class of charging infrastructure enables the differentiation of selected features, e.g. by the used rollout strategy. Overall, the paper demonstrates the application of statistical methodologies to energy data and provides insights on factors potentially shaping the energy consumption that could be utilized when developing models to inform charging infrastructure deployment and planning of power grids.

Milan Straka, Pasquale De Falco, Gabriella Ferruzzi et al.

The availability of charging infrastructure is essential for large-scale adoption of electric vehicles (EV). Charging patterns and the utilization of infrastructure have consequences not only for the energy demand, loading local power grids but influence the economic returns, parking policies and further adoption of EVs. We develop a data-driven approach that is exploiting predictors compiled from GIS data describing the urban context and urban activities near charging infrastructure to explore correlations with a comprehensive set of indicators measuring the performance of charging infrastructure. The best fit was identified for the size of the unique group of visitors (popularity) attracted by the charging infrastructure. Consecutively, charging infrastructure is ranked by popularity. The question of whether or not a given charging spot belongs to the top tier is posed as a binary classification problem and predictive performance of logistic regression regularized with an l-1 penalty, random forests and gradient boosted regression trees is evaluated. Obtained results indicate that the collected predictors contain information that can be used to predict the popularity of charging infrastructure. The significance of predictors and how they are linked with the popularity are explored as well. The proposed methodology can be used to inform charging infrastructure deployment strategies.

Jakub Náplava, Milan Straka

In this paper, we describe our systems submitted to the Building Educational Applications (BEA) 2019 Shared Task (Bryant et al., 2019). We participated in all three tracks. Our models are NMT systems based on the Transformer model, which we improve by incorporating several enhancements: applying dropout to whole source and target words, weighting target subwords, averaging model checkpoints, and using the trained model iteratively for correcting the intermediate translations. The system in the Restricted Track is trained on the provided corpora with oversampled "cleaner" sentences and reaches 59.39 F0.5 score on the test set. The system in the Low-Resource Track is trained from Wikipedia revision histories and reaches 44.13 F0.5 score. Finally, we finetune the system from the Low-Resource Track on restricted data and achieve 64.55 F0.5 score, placing third in the Unrestricted Track.

Milan Straka, Jana Straková, Jan Hajič

Contextualized embeddings, which capture appropriate word meaning depending on context, have recently been proposed. We evaluate two meth ods for precomputing such embeddings, BERT and Flair, on four Czech text processing tasks: part-of-speech (POS) tagging, lemmatization, dependency pars ing and named entity recognition (NER). The first three tasks, POS tagging, lemmatization and dependency parsing, are evaluated on two corpora: the Prague Dependency Treebank 3.5 and the Universal Dependencies 2.3. The named entity recognition (NER) is evaluated on the Czech Named Entity Corpus 1.1 and 2.0. We report state-of-the-art results for the above mentioned tasks and corpora.

Milan Straka, Jana Straková, Jan Hajič

We present an extensive evaluation of three recently proposed methods for contextualized embeddings on 89 corpora in 54 languages of the Universal Dependencies 2.3 in three tasks: POS tagging, lemmatization, and dependency parsing. Employing the BERT, Flair and ELMo as pretrained embedding inputs in a strong baseline of UDPipe 2.0, one of the best-performing systems of the CoNLL 2018 Shared Task and an overall winner of the EPE 2018, we present a one-to-one comparison of the three contextualized word embedding methods, as well as a comparison with word2vec-like pretrained embeddings and with end-to-end character-level word embeddings. We report state-of-the-art results in all three tasks as compared to results on UD 2.2 in the CoNLL 2018 Shared Task.

Milan Straka, Jana Straková, Jan Hajič

We present our contribution to the SIGMORPHON 2019 Shared Task: Crosslinguality and Context in Morphology, Task 2: contextual morphological analysis and lemmatization. We submitted a modification of the UDPipe 2.0, one of best-performing systems of the CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies and an overall winner of the The 2018 Shared Task on Extrinsic Parser Evaluation. As our first improvement, we use the pretrained contextualized embeddings (BERT) as additional inputs to the network; secondly, we use individual morphological features as regularization; and finally, we merge the selected corpora of the same language. In the lemmatization task, our system exceeds all the submitted systems by a wide margin with lemmatization accuracy 95.78 (second best was 95.00, third 94.46). In the morphological analysis, our system placed tightly second: our morphological analysis accuracy was 93.19, the winning system's 93.23.

Jana Straková, Milan Straka, Jan Hajič

We propose two neural network architectures for nested named entity recognition (NER), a setting in which named entities may overlap and also be labeled with more than one label. We encode the nested labels using a linearized scheme. In our first proposed approach, the nested labels are modeled as multilabels corresponding to the Cartesian product of the nested labels in a standard LSTM-CRF architecture. In the second one, the nested NER is viewed as a sequence-to-sequence problem, in which the input sequence consists of the tokens and output sequence of the labels, using hard attention on the word whose label is being predicted. The proposed methods outperform the nested NER state of the art on four corpora: ACE-2004, ACE-2005, GENIA and Czech CNEC. We also enrich our architectures with the recently published contextual embeddings: ELMo, BERT and Flair, reaching further improvements for the four nested entity corpora. In addition, we report flat NER state-of-the-art results for CoNLL-2002 Dutch and Spanish and for CoNLL-2003 English.

Daniel Kondratyuk, Tomáš Gavenčiak, Milan Straka et al.

We present LemmaTag, a featureless neural network architecture that jointly generates part-of-speech tags and lemmas for sentences by using bidirectional RNNs with character-level and word-level embeddings. We demonstrate that both tasks benefit from sharing the encoding part of the network, predicting tag subcategories, and using the tagger output as an input to the lemmatizer. We evaluate our model across several languages with complex morphology, which surpasses state-of-the-art accuracy in both part-of-speech tagging and lemmatization in Czech, German, and Arabic.