Collecting texts from the web enables a rapid creation of monolingual and parallel corpora of unprecedented size. However, unlike manually-collected corpora, authors and end users do not know which texts make up the web collections. In this work, we analyse the content of seven European parallel web corpora, collected from national top-level domains, by analysing the English variety and genre distribution in them. We develop and provide a lexicon-based British-American variety classifier, which we use to identify the English variety. In addition, we apply a Transformer-based genre classifier to corpora to analyse genre distribution and the interplay between genres and English varieties. The results reveal significant differences among the seven corpora in terms of different genre distribution and different preference for English varieties.

Automatic discrimination between Bosnian, Croatian, Montenegrin and Serbian is a hard task due to the mutual intelligibility of these South-Slavic languages. In this paper, we introduce the BENCHić-lang benchmark for discriminating between these four languages. The benchmark consists of two datasets from different domains - a Twitter and a news dataset - selected with the aim of fostering cross-dataset evaluation of different modelling approaches. We experiment with the baseline SVM models, based on character n-grams, which perform nicely in-dataset, but do not generalize well in cross-dataset experiments. Thus, we introduce another approach, exploiting only web-crawled data and the weak supervision signal coming from the respective country/language top-level domains. The resulting simple Naive Bayes model, based on less than a thousand word features extracted from web data, outperforms the baseline models in the cross-dataset scenario and achieves good levels of generalization across datasets.

This report presents the results of the shared tasks organized as part of the VarDial Evaluation Campaign 2023. The campaign is part of the tenth workshop on Natural Language Processing (NLP) for Similar Languages, Varieties and Dialects (VarDial), co-located with EACL 2023. Three separate shared tasks were included this year: Slot and intent detection for low-resource language varieties (SID4LR), Discriminating Between Similar Languages – True Labels (DSL-TL), and Discriminating Between Similar Languages – Speech (DSL-S). All three tasks were organized for the first time this year.

We present the most relevant results of the project MaCoCu: Massive collection and curation of monolingual and bilingual data: focus on under-resourced languages in its second year. To date, parallel and monolingual corpora have been produced for seven low-resourced European languages by crawling large amounts of textual data from selected top-level domains of the Internet; both human and automatic evaluation show its usefulness. In addition, several large language models pretrained on MaCoCu data have been published, as well as the code used to collect and curate the data.

We present version 1.3 of the PARSEME multilingual corpus annotated with verbal multiword expressions. Since the previous version, new languages have joined the undertaking of creating such a resource, some of the already existing corpora have been enriched with new annotated texts, while others have been enhanced in various ways. The PARSEME multilingual corpus represents 26 languages now. All monolingual corpora therein use Universal Dependencies v.2 tagset. They are (re-)split observing the PARSEME v.1.2 standard, which puts impact on unseen VMWEs. With the current iteration, the corpus release process has been detached from shared tasks; instead, a process for continuous improvement and systematic releases has been introduced.

This paper presents the creation and evaluation of a new version of the reference SSJ Universal Dependencies Treebank for Slovenian, which has been substantially improved and extended to almost double the original size. The process was based on the initial revision and documentation of the language-specific UD annotation guidelines for Slovenian and the corresponding modification of the original SSJ annotations, followed by a two-stage annotation campaign, in which two new subsets have been added, the previously unreleased sentences from the ssj500k corpus and the Slovenian subset of the ELEXIS parallel corpus. The annotation campaign resulted in an extended version of the SSJ UD treebank with 5,435 newly added sentences comprising of 126,427 tokens. To evaluate the potential benefits of this data increase for Slovenian dependency parsing, we compared the performance of the classla-stanza dependency parser trained on the old and the new SSJ data when evaluated on the new SSJ test set and its subsets. Our results show an increase of LAS performance in general, especially for previously under-represented syntactic phenomena, such as lists, elliptical constructions and appositions, but also confirm the distinct nature of the two newly added subsets and the diversification of the SSJ treebank as a whole.

This paper presents a new training dataset for automatic genre identification GINCO, which is based on 1,125 crawled Slovenian web documents that consist of 650,000 words. Each document was manually annotated for genre with a new annotation schema that builds upon existing schemata, having primarily clarity of labels and inter-annotator agreement in mind. The dataset consists of various challenges related to web-based data, such as machine translated content, encoding errors, multiple contents presented in one document etc., enabling evaluation of classifiers in realistic conditions. The initial machine learning experiments on the dataset show that (1) pre-Transformer models are drastically less able to model the phenomena, with macro F1 metrics ranging around 0.22, while Transformer-based models achieve scores of around 0.58, and (2) multilingual Transformer models work as well on the task as the monolingual models that were previously proven to be superior to multilingual models on standard NLP tasks.

We introduce the project “MaCoCu: Massive collection and curation of monolingual and bilingual data: focus on under-resourced languages”, funded by the Connecting Europe Facility, which is aimed at building monolingual and parallel corpora for under-resourced European languages. The approach followed consists of crawling large amounts of textual data from carefully selected top-level domains of the Internet, and then applying a curation and enrichment pipeline. In addition to corpora, the project will release successive versions of the free/open-source web crawling and curation software used.

In ParlaMint I, a CLARIN-ERIC supported project in pandemic times, a set of comparable and uniformly annotated multilingual corpora for 17 national parliaments were developed and released in 2021. For 2022 and 2023, the project has been extended to ParlaMint II, again with the CLARIN ERIC financial support, in order to enhance the existing corpora with new data and metadata; upgrade the XML schema; add corpora for 10 new parliaments; provide more application scenarios and carry out additional experiments. The paper reports on these planned steps, including some that have already been taken, and outlines future plans.

This paper presents our bootstrapping efforts of producing the first large freely available Croatian automatic speech recognition (ASR) dataset, 1,816 hours in size, obtained from parliamentary transcripts and recordings from the ParlaMint corpus. The bootstrapping approach to the dataset building relies on a commercial ASR system for initial data alignment, and building a multilingual-transformer-based ASR system from the initial data for full data alignment. Experiments on the resulting dataset show that the difference between the spoken content and the parliamentary transcripts is present in ~4-5% of words, which is also the word error rate of our best-performing ASR system. Interestingly, fine-tuning transformer models on either normalized or original data does not show a difference in performance. Models pre-trained on a subset of raw speech data consisting of Slavic languages only show to perform better than those pre-trained on a wider set of languages. With our public release of data, models and code, we are paving the way forward for the preparation of the multi-modal corpus of Croatian parliamentary proceedings, as well as for the development of similar free datasets, models and corpora for other under-resourced languages.

There is a shortage of high-quality corpora for South-Slavic languages. Such corpora are useful to computer scientists and researchers in social sciences and humanities alike, focusing on numerous linguistic, content analysis, and natural language processing applications. This paper presents a workflow for mining Wikipedia content and processing it into linguistically-processed corpora, applied on the Bosnian, Bulgarian, Croatian, Macedonian, Serbian, Serbo-Croatian and Slovenian Wikipedia. We make the resulting seven corpora publicly available. We showcase these corpora by comparing the content of the underlying Wikipedias, our assumption being that the content of the Wikipedias reflects broadly the interests in various topics in these Balkan nations. We perform the content comparison by using topic modelling algorithms and various distribution comparisons. The results show that all Wikipedias are topically rather similar, with all of them covering art, culture, and literature, whereas they contain differences in geography, politics, history and science.

In this paper, we describe experiments designed to evaluate the impact of stylometric and emotion-based features on hate speech detection: the task of classifying textual content into hate or non-hate speech classes. Our experiments are conducted for three languages – English, Slovene, and Dutch – both in in-domain and cross-domain setups, and aim to investigate hate speech using features that model two linguistic phenomena: the writing style of hateful social media content operationalized as function word usage on the one hand, and emotion expression in hateful messages on the other hand. The results of experiments with features that model different combinations of these phenomena support our hypothesis that stylometric and emotion-based features are robust indicators of hate speech. Their contribution remains persistent with respect to domain and language variation. We show that the combination of features that model the targeted phenomena outperforms words and character n-gram features under cross-domain conditions, and provides a significant boost to deep learning models, which currently obtain the best results, when combined with them in an ensemble.

In this paper we describe a transformer model pre-trained on 8 billion tokens of crawled text from the Croatian, Bosnian, Serbian and Montenegrin web domains. We evaluate the transformer model on the tasks of part-of-speech tagging, named-entity-recognition, geo-location prediction and commonsense causal reasoning, showing improvements on all tasks over state-of-the-art models. For commonsense reasoning evaluation we introduce COPA-HR - a translation of the Choice of Plausible Alternatives (COPA) dataset into Croatian. The BERTić model is made available for free usage and further task-specific fine-tuning through HuggingFace.

This paper describes the results of the shared tasks organized as part of the VarDial Evaluation Campaign 2021. The campaign was part of the eighth workshop on Natural Language Processing (NLP) for Similar Languages, Varieties and Dialects (VarDial), co-located with EACL 2021. Four separate shared tasks were included this year: Dravidian Language Identification (DLI), Romanian Dialect Identification (RDI), Social Media Variety Geolocation (SMG), and Uralic Language Identification (ULI). DLI was organized for the first time and the other three continued a series of tasks from previous evaluation campaigns.

This paper describes the Helsinki–Ljubljana contribution to the VarDial 2021 shared task on social media variety geolocation. Following our successful participation at VarDial 2020, we again propose constrained and unconstrained systems based on the BERT architecture. In this paper, we report experiments with different tokenization settings and different pre-trained models, and we contrast our parameter-free regression approach with various classification schemes proposed by other participants at VarDial 2020. Both the code and the best-performing pre-trained models are made freely available.

This paper describes the HEL-LJU submissions to the MultiLexNorm shared task on multilingual lexical normalization. Our system is based on a BERT token classification preprocessing step, where for each token the type of the necessary transformation is predicted (none, uppercase, lowercase, capitalize, modify), and a character-level SMT step where the text is translated from original to normalized given the BERT-predicted transformation constraints. For some languages, depending on the results on development data, the training data was extended by back-translating OpenSubtitles data. In the final ordering of the ten participating teams, the HEL-LJU team has taken the second place, scoring better than the previous state-of-the-art.

Lexical normalization is the task of transforming an utterance into its standardized form. This task is beneficial for downstream analysis, as it provides a way to harmonize (often spontaneous) linguistic variation. Such variation is typical for social media on which information is shared in a multitude of ways, including diverse languages and code-switching. Since the seminal work of Han and Baldwin (2011) a decade ago, lexical normalization has attracted attention in English and multiple other languages. However, there exists a lack of a common benchmark for comparison of systems across languages with a homogeneous data and evaluation setup. The MultiLexNorm shared task sets out to fill this gap. We provide the largest publicly available multilingual lexical normalization benchmark including 13 language variants. We propose a homogenized evaluation setup with both intrinsic and extrinsic evaluation. As extrinsic evaluation, we use dependency parsing and part-of-speech tagging with adapted evaluation metrics (a-LAS, a-UAS, and a-POS) to account for alignment discrepancies. The shared task hosted at W-NUT 2021 attracted 9 participants and 18 submissions. The results show that neural normalization systems outperform the previous state-of-the-art system by a large margin. Downstream parsing and part-of-speech tagging performance is positively affected but to varying degrees, with improvements of up to 1.72 a-LAS, 0.85 a-UAS, and 1.54 a-POS for the winning system.

This paper presents the results of the news translation task and the similar language translation task, both organised alongside the Conference on Machine Translation (WMT) 2020. In the news task, participants were asked to build machine translation systems for any of 11 language pairs, to be evaluated on test sets consisting mainly of news stories. The task was also opened up to additional test suites to probe specific aspects of translation. In the similar language translation task, participants built machine translation systems for translating between closely related pairs of languages.

This paper presents the results of the VarDial Evaluation Campaign 2020 organized as part of the seventh workshop on Natural Language Processing (NLP) for Similar Languages, Varieties and Dialects (VarDial), co-located with COLING 2020. The campaign included three shared tasks each focusing on a different challenge of language and dialect identification: Romanian Dialect Identification (RDI), Social Media Variety Geolocation (SMG), and Uralic Language Identification (ULI). The campaign attracted 30 teams who enrolled to participate in one or multiple shared tasks and 14 of them submitted runs across the three shared tasks. Finally, 11 papers describing participating systems are published in the VarDial proceedings and referred to in this report.

This paper describes the Helsinki-Ljubljana contribution to the VarDial shared task on social media variety geolocation. Our solutions are based on the BERT Transformer models, the constrained versions of our models reaching 1st place in two subtasks and 3rd place in one subtask, while our unconstrained models outperform all the constrained systems by a large margin. We show in our analyses that Transformer-based models outperform traditional models by far, and that improvements obtained by pre-training models on large quantities of (mostly standard) text are significant, but not drastic, with single-language models also outperforming multilingual models. Our manual analysis shows that two types of signals are the most crucial for a (mis)prediction: named entities and dialectal features, both of which are handled well by our models.

In this paper, we present emotion lexicons of Croatian, Dutch and Slovene, based on manually corrected automatic translations of the English NRC Emotion lexicon. We evaluate the impact of the translation changes by measuring the change in supervised classification results of socially unacceptable utterances when lexicon information is used for feature construction. We further showcase the usage of the lexicons by calculating the difference in emotion distributions in texts containing and not containing socially unacceptable discourse, comparing them across four languages (English, Croatian, Dutch, Slovene) and two topics (migrants and LGBT). We show significant and consistent improvements in automatic classification across all languages and topics, as well as consistent (and expected) emotion distributions across all languages and topics, proving for the manually corrected lexicons to be a useful addition to the severely lacking area of emotion lexicons, the crucial resource for emotive analysis of text.

We describe a new version of the Gigafida reference corpus of Slovene. In addition to updating the corpus with new material and annotating it with better tools, the focus of the upgrade was also on its transformation from a general reference corpus, which contains all language variants including non-standard language, to the corpus of standard (written) Slovene. This decision could be implemented as new corpora dedicated specifically to non-standard language emerged recently. In the new version, the whole Gigafida corpus was deduplicated for the first time, which facilitates automatic extraction of data for the purposes of compilation of new lexicographic resources such as the collocations dictionary and the thesaurus of Slovene.

State of the art natural language processing tools are built on context-dependent word embeddings, but no direct method for evaluating these representations currently exists. Standard tasks and datasets for intrinsic evaluation of embeddings are based on judgements of similarity, but ignore context; standard tasks for word sense disambiguation take account of context but do not provide continuous measures of meaning similarity. This paper describes an effort to build a new dataset, CoSimLex, intended to fill this gap. Building on the standard pairwise similarity task of SimLex-999, it provides context-dependent similarity measures; covers not only discrete differences in word sense but more subtle, graded changes in meaning; and covers not only a well-resourced language (English) but a number of less-resourced languages. We define the task and evaluation metrics, outline the dataset collection methodology, and describe the status of the dataset so far.

This paper presents the Graded Word Similarity in Context (GWSC) task which asked participants to predict the effects of context on human perception of similarity in English, Croatian, Slovene and Finnish. We received 15 submissions and 11 system description papers. A new dataset (CoSimLex) was created for evaluation in this task: it contains pairs of words, each annotated within two different contexts. Systems beat the baselines by significant margins, but few did well in more than one language or subtask. Almost every system employed a Transformer model, but with many variations in the details: WordNet sense embeddings, translation of contexts, TF-IDF weightings, and the automatic creation of datasets for fine-tuning were all used to good effect.

We present experiments on Slovenian, Croatian and Serbian morphosyntactic annotation and lemmatisation between the former state-of-the-art for these three languages and one of the best performing systems at the CoNLL 2018 shared task, the Stanford NLP neural pipeline. Our experiments show significant improvements in morphosyntactic annotation, especially on categories where either semantic knowledge is needed, available through word embeddings, or where long-range dependencies have to be modelled. On the other hand, on the task of lemmatisation no improvements are obtained with the neural solution, mostly due to the heavy dependence of the task on the lookup in an external lexicon, but also due to obvious room for improvements in the Stanford NLP pipeline’s lemmatisation.

The notions of concreteness and imageability, traditionally important in psycholinguistics, are gaining significance in semantic-oriented natural language processing tasks. In this paper we investigate the predictability of these two concepts via supervised learning, using word embeddings as explanatory variables. We perform predictions both within and across languages by exploiting collections of cross-lingual embeddings aligned to a single vector space. We show that the notions of concreteness and imageability are highly predictable both within and across languages, with a moderate loss of up to 20% in correlation when predicting across languages. We further show that the cross-lingual transfer via word embeddings is more efficient than the simple transfer via bilingual dictionaries.

We present the results and the findings of the Second VarDial Evaluation Campaign on Natural Language Processing (NLP) for Similar Languages, Varieties and Dialects. The campaign was organized as part of the fifth edition of the VarDial workshop, collocated with COLING’2018. This year, the campaign included five shared tasks, including two task re-runs – Arabic Dialect Identification (ADI) and German Dialect Identification (GDI) –, and three new tasks – Morphosyntactic Tagging of Tweets (MTT), Discriminating between Dutch and Flemish in Subtitles (DFS), and Indo-Aryan Language Identification (ILI). A total of 24 teams submitted runs across the five shared tasks, and contributed 22 system description papers, which were included in the VarDial workshop proceedings and are referred to in this report.

This paper presents two systems taking part in the Morphosyntactic Tagging of Tweets shared task on Slovene, Croatian and Serbian data, organized inside the VarDial Evaluation Campaign. While one system relies on the traditional method for sequence labeling (conditional random fields), the other relies on its neural alternative (bidirectional long short-term memory). We investigate the similarities and differences of these two approaches, showing that both methods yield very good and quite similar results, with the neural model outperforming the traditional one more as the level of non-standardness of the text increases. Through an error analysis we show that the neural system is better at long-range dependencies, while the traditional system excels and slightly outperforms the neural system at the local ones. We present in the paper new state-of-the-art results in morphosyntactic annotation of non-standard text for Slovene, Croatian and Serbian.

This paper presents two large newly constructed datasets of moderated news comments from two highly popular online news portals in the respective countries: the Slovene RTV MCC and the Croatian 24sata. The datasets are analyzed by performing manual annotation of the types of the content which have been deleted by moderators and by investigating deletion trends among users and threads. Next, initial experiments on automatically detecting the deleted content in the datasets are presented. Both datasets are published in encrypted form, to enable others to perform experiments on detecting content to be deleted without revealing potentially inappropriate content. Finally, the baseline classification models trained on the non-encrypted datasets are disseminated as well to enable real-world use.

Gender prediction has typically focused on lexical and social network features, yielding good performance, but making systems highly language-, topic-, and platform dependent. Cross-lingual embeddings circumvent some of these limitations, but capture gender-specific style less. We propose an alternative: bleaching text, i.e., transforming lexical strings into more abstract features. This study provides evidence that such features allow for better transfer across languages. Moreover, we present a first study on the ability of humans to perform cross-lingual gender prediction. We find that human predictive power proves similar to that of our bleached models, and both perform better than lexical models.

We present the results of the VarDial Evaluation Campaign on Natural Language Processing (NLP) for Similar Languages, Varieties and Dialects, which we organized as part of the fourth edition of the VarDial workshop at EACL’2017. This year, we included four shared tasks: Discriminating between Similar Languages (DSL), Arabic Dialect Identification (ADI), German Dialect Identification (GDI), and Cross-lingual Dependency Parsing (CLP). A total of 19 teams submitted runs across the four tasks, and 15 of them wrote system description papers.

The paper documents the procedure of building a new Universal Dependencies (UDv2) treebank for Serbian starting from an existing Croatian UDv1 treebank and taking into account the other Slavic UD annotation guidelines. We describe the automatic and manual annotation procedures, discuss the annotation of Slavic-specific categories (case governing quantifiers, reflexive pronouns, question particles) and propose an approach to handling deverbal nouns in Slavic languages.

In this paper we present the adaptations of a state-of-the-art tagger for South Slavic languages to non-standard texts on the example of the Slovene language. We investigate the impact of introducing in-domain training data as well as additional supervision through external resources or tools like word clusters and word normalization. We remove more than half of the error of the standard tagger when applied to non-standard texts by training it on a combination of standard and non-standard training data, while enriching the data representation with external resources removes additional 11 percent of the error. The final configuration achieves tagging accuracy of 87.41% on the full morphosyntactic description, which is, nevertheless, still quite far from the accuracy of 94.27% achieved on standard text.

In this paper we present a set of experiments and analyses on predicting the gender of Twitter users based on language-independent features extracted either from the text or the metadata of users’ tweets. We perform our experiments on the TwiSty dataset containing manual gender annotations for users speaking six different languages. Our classification results show that, while the prediction model based on language-independent features performs worse than the bag-of-words model when training and testing on the same language, it regularly outperforms the bag-of-words model when applied to different languages, showing very stable results across various languages. Finally we perform a comparative analysis of feature effect sizes across the six languages and show that differences in our features correspond to cultural distances.

In this paper we present the legal framework, dataset and annotation schema of socially unacceptable discourse practices on social networking platforms in Slovenia. On this basis we aim to train an automatic identification and classification system with which we wish contribute towards an improved methodology, understanding and treatment of such practices in the contemporary, increasingly multicultural information society.

In this paper we present a tagger developed for inflectionally rich languages for which both a training corpus and a lexicon are available. We do not constrain the tagger by the lexicon entries, allowing both for lexicon incompleteness and noisiness. By using the lexicon indirectly through features we allow for known and unknown words to be tagged in the same manner. We test our tagger on Slovene data, obtaining a 25% error reduction of the best previous results both on known and unknown words. Given that Slovene is, in comparison to some other Slavic languages, a well-resourced language, we perform experiments on the impact of token (corpus) vs. type (lexicon) supervision, obtaining useful insights in how to balance the effort of extending resources to yield better tagging results.

This paper presents an approach for building large monolingual corpora and, at the same time, extracting parallel data by crawling the top-level domain of a given language of interest. For gathering linguistically relevant data from top-level domains we use the SpiderLing crawler, modified to crawl data written in multiple languages. The output of this process is then fed to Bitextor, a tool for harvesting parallel data from a collection of documents. We call the system combining these two tools Spidextor, a blend of the names of its two crucial parts. We evaluate the described approach intrinsically by measuring the accuracy of the extracted bitexts from the Croatian top-level domain “.hr” and the Slovene top-level domain “.si”, and extrinsically on the English-Croatian language pair by comparing an SMT system built from the crawled data with third-party systems. We finally present parallel datasets collected with our approach for the English-Croatian, English-Finnish, English-Serbian and English-Slovene language pairs.

In the paper authors present the Croatian corpus of non-professional written language. Consisting of two subcorpora, i.e. the clinical subcorpus, consisting of written texts produced by speakers with various types of language disorders, and the healthy speakers subcorpus, as well as by the levels of its annotation, it offers an opportunity for different lines of research. The authors present the corpus structure, describe the sampling methodology, explain the levels of annotation, and give some very basic statistics. On the basis of data from the corpus, existing language technologies for Croatian are adapted in order to be implemented in a platform facilitating text production to speakers with language disorders. In this respect, several analyses of the corpus data and a basic evaluation of the developed technologies are presented.

In computer-mediated communication, Latin-based scripts users often omit diacritics when writing. Such text is typically easily understandable to humans but very difficult for computational processing because many words become ambiguous or unknown. Letter-level approaches to diacritic restoration generalise better and do not require a lot of training data but word-level approaches tend to yield better results. However, they typically rely on a lexicon which is an expensive resource, not covering non-standard forms, and often not available for less-resourced languages. In this paper we present diacritic restoration models that are trained on easy-to-acquire corpora. We test three different types of corpora (Wikipedia, general web, Twitter) for three South Slavic languages (Croatian, Serbian and Slovene) and evaluate them on two types of text: standard (Wikipedia) and non-standard (Twitter). The proposed approach considerably outperforms charlifter, so far the only open source tool available for this task. We make the best performing systems freely available.

In this paper we present newly developed inflectional lexcions and manually annotated corpora of Croatian and Serbian. We introduce hrLex and srLex - two freely available inflectional lexicons of Croatian and Serbian - and describe the process of building these lexicons, supported by supervised machine learning techniques for lemma and paradigm prediction. Furthermore, we introduce hr500k, a manually annotated corpus of Croatian, 500 thousand tokens in size. We showcase the three newly developed resources on the task of morphosyntactic annotation of both languages by using a recently developed CRF tagger. We achieve best results yet reported on the task for both languages, beating the HunPos baseline trained on the same datasets by a wide margin.

In this paper we present a series of experiments on discriminating between private and corporate accounts on Twitter. We define features based on Twitter metadata, morphosyntactic tags and surface forms, showing that the simple bag-of-words model achieves single best results that can, however, be improved by building a weighted soft ensemble of classifiers based on each feature type. Investigating the time and language dependence of each feature type delivers quite unexpecting results showing that features based on metadata are neither time- nor language-insensitive as the way the two user groups use the social network varies heavily through time and space.

We present the results of the third edition of the Discriminating between Similar Languages (DSL) shared task, which was organized as part of the VarDial’2016 workshop at COLING’2016. The challenge offered two subtasks: subtask 1 focused on the identification of very similar languages and language varieties in newswire texts, whereas subtask 2 dealt with Arabic dialect identification in speech transcripts. A total of 37 teams registered to participate in the task, 24 teams submitted test results, and 20 teams also wrote system description papers. High-order character n-grams were the most successful feature, and the best classification approaches included traditional supervised learning methods such as SVM, logistic regression, and language models, while deep learning approaches did not perform very well.

Massive Open Online Courses have been growing rapidly in size and impact. Yet the language barrier constitutes a major growth impediment in reaching out all people and educating all citizens. A vast majority of educational material is available only in English, and state-of-the-art machine translation systems still have not been tailored for this peculiar genre. In addition, a mere collection of appropriate in-domain training material is a challenging task. In this work, we investigate statistical machine translation of lecture subtitles from English into Croatian, which is morphologically rich and generally weakly supported, especially for the educational domain. We show that results comparable with publicly available systems trained on much larger data can be achieved if a small in-domain training set is used in combination with additional in-domain corpus originating from the closely related Serbian language.

In this paper we present a newly developed tool that enables researchers interested in spatial variation of language to define a geographic perimeter of interest, collect data from the Twitter streaming API published in that perimeter, filter the obtained data by language and country, define and extract variables of interest and analyse the extracted variables by one spatial statistic and two spatial visualisations. We showcase the tool on the area and a selection of languages spoken in former Yugoslavia. By defining the perimeter, languages and a series of linguistic variables of interest we demonstrate the data collection, processing and analysis capabilities of the tool.

In this paper we compare two tools for automatically harvesting bitexts from multilingual websites: bitextor and ILSP-FC. We used both tools for crawling 21 multilingual websites from the tourism domain to build a domain-specific English―Croatian parallel corpus. Different settings were tried for both tools and 10,662 unique document pairs were obtained. A sample of about 10% of them was manually examined and the success rate was computed on the collection of pairs of documents detected by each setting. We compare the performance of the settings and the amount of different corpora detected by each setting. In addition, we describe the resource obtained, both by the settings and through the human evaluation, which has been released as a high-quality parallel corpus.

We present SETimes.HR ― the first linguistically annotated corpus of Croatian that is freely available for all purposes. The corpus is built on top of the SETimes parallel corpus of nine Southeast European languages and English. It is manually annotated for lemmas, morphosyntactic tags, named entities and dependency syntax. We couple the corpus with domain-sensitive test sets for Croatian and Serbian to support direct model transfer evaluation between these closely related languages. We build and evaluate statistical models for lemmatization, morphosyntactic tagging, named entity recognition and dependency parsing on top of SETimes.HR and the test sets, providing the state of the art in all the tasks. We make all resources presented in the paper freely available under a very permissive licensing scheme.

This paper presents a novel approach for parallel data generation using machine translation and quality estimation. Our study focuses on pivot-based machine translation from English to Croatian through Slovene. We generate an English―Croatian version of the Europarl parallel corpus based on the English―Slovene Europarl corpus and the Apertium rule-based translation system for Slovene―Croatian. These experiments are to be considered as a first step towards the generation of reliable synthetic parallel data for under-resourced languages. We first collect small amounts of aligned parallel data for the Slovene―Croatian language pair in order to build a quality estimation system for sentence-level Translation Edit Rate (TER) estimation. We then infer TER scores on automatically translated Slovene to Croatian sentences and use the best translations to build an English―Croatian statistical MT system. We show significant improvement in terms of automatic metrics obtained on two test sets using our approach compared to a random selection of synthetic parallel data.

This paper presents TweetCaT, an open-source Python tool for building Twitter corpora that was designed for smaller languages. Using the Twitter search API and a set of seed terms, the tool identifies users tweeting in the language of interest together with their friends and followers. By running the tool for 235 days we tested it on the task of collecting two monitor corpora, one for Croatian and Serbian and the other for Slovene, thus also creating new and valuable resources for these languages. A post-processing step on the collected corpus is also described, which filters out users that tweet predominantly in a foreign language thus further cleans the collected corpora. Finally, an experiment on discriminating between Croatian and Serbian Twitter users is reported.

In this paper we present the construction process of a web corpus of Catalan built from the content of the .cat top-level domain. For collecting and processing data we use the Brno pipeline with the spiderling crawler and its accompanying tools. To the best of our knowledge the corpus represents the largest existing corpus of Catalan containing 687 million words, which is a significant increase given that until now the biggest corpus of Catalan, CuCWeb, counts 166 million words. We evaluate the resulting resource on the tasks of language modeling and statistical machine translation (SMT) by calculating LM perplexity and incorporating the LM in the SMT pipeline. We compare language models trained on different subsets of the resource with those trained on the Catalan Wikipedia and the target side of the parallel data used to train the SMT system.

This paper presents an approach to extract translation equivalents from comparable corpora for polysemous nouns. As opposed to the standard approaches that build a single context vector for all occurrences of a given headword, we first disambiguate the headword with third-party sense taggers and then build a separate context vector for each sense of the headword. Since state-of-the-art word sense disambiguation tools are still far from perfect, we also tried to improve the results by combining the sense assignments provided by two different sense taggers. Evaluation of the results shows that we outperform the baseline (0.473) in all the settings we experimented with, even when using only one sense tagger, and that the best-performing results are indeed obtained by taking into account the intersection of both sense taggers (0.720).

This paper describes the process of building a newspaper corpus annotated with events described in specific documents. The main difference to the corpora built as part of the TDT initiative is that documents are not annotated by topics, but by specific events they describe. Additionally, documents are gathered from sixteen sources and all documents in the corpus are annotated with the corresponding event. The annotation process consists of a browsing and a searching step. Experiments are performed with a threshold that could be used in the browsing step yielding the result of having to browse through only 1% of document pairs for a 2% loss of relevant document pairs. A statistical analysis of the annotated corpus is undertaken showing that most events are described by few documents while just some events are reported by many documents. The inter-annotator agreement measures show high agreement concerning grouping documents into event clusters, but show a much lower agreement concerning the number of events the documents are organized into. An initial experiment is described giving a baseline for further research on this corpus.

The paper presents results of an experiment dealing with sentiment analysis of Croatian text from the domain of finance. The goal of the experiment was to design a system model for automatic detection of general sentiment and polarity phrases in these texts. We have assembled a document collection from web sources writing on the financial market in Croatia and manually annotated articles from a subset of that collection for general sentiment. Additionally, we have manually annotated a number of these articles for phrases encoding positive or negative sentiment within a text. In the paper, we provide an analysis of the compiled resources. We show a statistically significant correspondence (1) between the overall market trend on the Zagreb Stock Exchange and the number of positively and negatively accented articles within periods of trend and (2) between the general sentiment of articles and the number of polarity phrases within those articles. We use this analysis as an input for designing a rule-based local grammar system for automatic detection of polarity phrases and evaluate it on held out data. The system achieves F1-scores of 0.61 (P: 0.94, R: 0.45) and 0.63 (P: 0.97, R: 0.47) on positive and negative polarity phrases.

This paper describes methods used for generating a morphological lexicon of organization entity names in Croatian. This resource is intended for two primary tasks: template-based natural language generation and named entity identification. The main problems concerning the lexicon generation are high level of inflection in Croatian and low linguistic quality of the primary resource containing named entities in normal form. The problem is divided into two subproblems concerning single-word and multi-word expressions. The single-word problem is solved by training a supervised learning algorithm called linear successive abstraction. With existing common language morphological resources and two simple hand-crafted rules backing up the algorithm, accuracy of 98.70% on the test set is achieved. The multi-word problem is solved through a semi-automated process for multi-word entities occurring in the first 10,000 named entities. The generated multi-word lexicon will be used for natural language generation only while named entity identification will be solved algorithmically in forthcoming research. The single-word lexicon is capable of handling both tasks.