Large language models (LLMs) excel in many tasks in NLP and beyond, but most open models have very limited coverage of smaller languages and LLM work tends to focus on languages where nearly unlimited data is available for pretraining. In this work, we study the challenges of creating LLMs for Finnish, a language spoken by less than 0.1% of the world population. We compile an extensive dataset of Finnish combining web crawls, news, social media and eBooks. We pursue two approaches to pretrain models: 1) we train seven monolingual models from scratch (186M to 13B parameters) dubbed FinGPT, 2) we continue the pretraining of the multilingual BLOOM model on a mix of its original training data and Finnish, resulting in a 176 billion parameter model we call BLUUMI. For model evaluation, we introduce FIN-bench, a version of BIG-bench with Finnish tasks. We also assess other model qualities such as toxicity and bias. Our models and tools are openly available at https://turkunlp.org/gpt3-finnish.

Relation Extraction (RE) remains a challenging task, especially when considering realistic out-of-domain evaluations. One of the main reasons for this is the limited training size of current RE datasets: obtaining high-quality (manually annotated) data is extremely expensive and cannot realistically be repeated for each new domain. An intermediate training step on data from related tasks has shown to be beneficial across many NLP tasks. However, this setup still requires supplementary annotated data, which is often not available. In this paper, we investigate intermediate pre-training specifically for RE. We exploit the affinity between syntactic structure and semantic RE, and identify the syntactic relations which are closely related to RE by being on the shortest dependency path between two entities. We then take advantage of the high accuracy of current syntactic parsers in order to automatically obtain large amounts of low-cost pre-training data. By pre-training our RE model on the relevant syntactic relations, we are able to outperform the baseline in five out of six cross-domain setups, without any additional annotated data.

Most research in Relation Extraction (RE) involves the English language, mainly due to the lack of multi-lingual resources. We propose Multi-CrossRE, the broadest multi-lingual dataset for RE, including 26 languages in addition to English, and covering six text domains. Multi-CrossRE is a machine translated version of CrossRE (Bassignana and Plank, 2022), with a sub-portion including more than 200 sentences in seven diverse languages checked by native speakers. We run a baseline model over the 26 new datasets and–as sanity check–over the 26 back-translations to English. Results on the back-translated data are consistent with the ones on the original English CrossRE, indicating high quality of the translation and the resulting dataset.

Due to the popularity of social media platforms and the sheer amount of user-generated content online, the automatic detection of toxic language has become crucial in the creation of a friendly and safe digital space. Previous work has been mostly focusing on English leaving many lower-resource languages behind. In this paper, we present novel resources for toxicity detection in Finnish by introducing two new datasets, a machine translated toxicity dataset for Finnish based on the widely used English Jigsaw dataset and a smaller test set of Suomi24 discussion forum comments originally written in Finnish and manually annotated following the definitions of the labels that were used to annotate the Jigsaw dataset. We show that machine translating the training data to Finnish provides better toxicity detection results than using the original English training data and zero-shot cross-lingual transfer with XLM-R, even with our newly annotated dataset from Suomi24.

In this paper, we describe a BERT model trained on the Eighteenth Century Collections Online (ECCO) dataset of digitized documents. The ECCO dataset poses unique modelling challenges due to the presence of Optical Character Recognition (OCR) artifacts. We establish the performance of the BERT model on a publication year prediction task against linear baseline models and human judgement, finding the BERT model to be superior to both and able to date the works, on average, with less than 7 years absolute error. We also explore how language change over time affects the model by analyzing the features the model uses for publication year predictions as given by the Integrated Gradients model explanation method.

Input saliency methods have recently become a popular tool for explaining predictions of deep learning models in NLP. Nevertheless, there has been little work investigating methods for aggregating prediction-level explanations to the class level, nor has a framework for evaluating such class explanations been established. We explore explanations based on XLM-R and the Integrated Gradients input attribution method, and propose 1) the Stable Attribution Class Explanation method (SACX) to extract keyword lists of classes in text classification tasks, and 2) a framework for the systematic evaluation of the keyword lists. We find that explanations of individual predictions are prone to noise, but that stable explanations can be effectively identified through repeated training and explanation. We evaluate on web register data and show that the class explanations are linguistically meaningful and distinguishing of the classes.

Evaluations in machine learning rarely use the latest metrics, datasets, or human evaluation in favor of remaining compatible with prior work. The compatibility, often facilitated through leaderboards, thus leads to outdated but standardized evaluation practices. We pose that the standardization is taking place in the wrong spot. Evaluation infrastructure should enable researchers to use the latest methods and what should be standardized instead is how to incorporate these new evaluation advances. We introduce GEMv2, the new version of the Generation, Evaluation, and Metrics Benchmark which uses a modular infrastructure for dataset, model, and metric developers to benefit from each other’s work. GEMv2 supports 40 documented datasets in 51 languages, ongoing online evaluation for all datasets, and our interactive tools make it easier to add new datasets to the living benchmark.

Automatic grouping of textual answers has the potential of allowing batch grading, but is challenging because the answers, especially longer essays, have many claims. To explore the feasibility of grouping together answers based on their semantic meaning, this paper investigates the grouping of short textual answers, proxies of single claims. This is approached as a paraphrase identification task, where neural and non-neural sentence embeddings and a paraphrase identification model are tested. These methods are evaluated on a dataset consisting of over 4000 short textual answers from various disciplines. The results map out the suitable question types for the paraphrase identification model and those for the neural and non-neural methods.

The prevailing practice in the academia is to evaluate the model performance on in-domain evaluation data typically set aside from the training corpus. However, in many real world applications the data on which the model is applied may very substantially differ from the characteristics of the training data. In this paper, we focus on Finnish out-of-domain parsing by introducing a novel UD Finnish-OOD out-of-domain treebank including five very distinct data sources (web documents, clinical, online discussions, tweets, and poetry), and a total of 19,382 syntactic words in 2,122 sentences released under the Universal Dependencies framework. Together with the new treebank, we present extensive out-of-domain parsing evaluation utilizing the available section-level information from three different Finnish UD treebanks (TDT, PUD, OOD). Compared to the previously existing treebanks, the new Finnish-OOD is shown include sections more challenging for the general parser, creating an interesting evaluation setting and yielding valuable information for those applying the parser outside of its training domain.

Deep neural language models such as BERT have enabled substantial recent advances in many natural language processing tasks. However, due to the effort and computational cost involved in their pre-training, such models are typically introduced only for a small number of high-resource languages such as English. While multilingual models covering large numbers of languages are available, recent work suggests monolingual training can produce better models, and our understanding of the tradeoffs between mono- and multilingual training is incomplete. In this paper, we introduce a simple, fully automated pipeline for creating language-specific BERT models from Wikipedia data and introduce 42 new such models, most for languages up to now lacking dedicated deep neural language models. We assess the merits of these models using cloze tests and the state-of-the-art UDify parser on Universal Dependencies data, contrasting performance with results using the multilingual BERT (mBERT) model. We find that the newly introduced WikiBERT models outperform mBERT in cloze tests for nearly all languages, and that UDify using WikiBERT models outperforms the parser using mBERT on average, with the language-specific models showing substantially improved performance for some languages, yet limited improvement or a decrease in performance for others. All of the methods and models introduced in this work are available under open licenses from https://github.com/turkunlp/wikibert.

We introduce a corpus with fine-grained named entity annotation for Finnish, following the OntoNotes guidelines to create a resource that is cross-lingually compatible with existing annotations for other languages. We combine and extend two NER corpora recently introduced for Finnish and revise their custom annotation scheme through a combination of automatic and manual processing steps. The resulting corpus consists of nearly 500,000 tokens annotated for over 50,000 mentions categorized into the 18 OntoNotes name and numeric entity types. We evaluate this resource and demonstrate its compatibility with the English OntoNotes annotations by training state-of-the-art mono-, bi- and multilingual deep learning models, finding both that the corpus allows highly accurate recognition of OntoNotes types at 93% F-score and that a comparable level of tagging accuracy can be achieved by a bilingual Finnish-English NER model.

In this paper, we introduce the first fully manually annotated paraphrase corpus for Finnish containing 53,572 paraphrase pairs harvested from alternative subtitles and news headings. Out of all paraphrase pairs in our corpus 98% are manually classified to be paraphrases at least in their given context, if not in all contexts. Additionally, we establish a manual candidate selection method and demonstrate its feasibility in high quality paraphrase selection in terms of both cost and quality.

This paper presents FISKMÖ, a project that focuses on the development of resources and tools for cross-linguistic research and machine translation between Finnish and Swedish. The goal of the project is the compilation of a massive parallel corpus out of translated material collected from web sources, public and private organisations and language service providers in Finland with its two official languages. The project also aims at the development of open and freely accessible translation services for those two languages for the general purpose and for domain-specific use. We have released new data sets with over 3 million translation units, a benchmark test set for MT development, pre-trained neural MT models with high coverage and competitive performance and a self-contained MT plugin for a popular CAT tool. The latter enables offline translation without dependencies on external services making it possible to work with highly sensitive data without compromising security concerns.

Universal Dependencies is an open community effort to create cross-linguistically consistent treebank annotation for many languages within a dependency-based lexicalist framework. The annotation consists in a linguistically motivated word segmentation; a morphological layer comprising lemmas, universal part-of-speech tags, and standardized morphological features; and a syntactic layer focusing on syntactic relations between predicates, arguments and modifiers. In this paper, we describe version 2 of the universal guidelines (UD v2), discuss the major changes from UD v1 to UD v2, and give an overview of the currently available treebanks for 90 languages.

We present the approach of the TurkuNLP group to the IWPT 2020 shared task on Multilingual Parsing into Enhanced Universal Dependencies. The task involves 28 treebanks in 17 different languages and requires parsers to generate graph structures extending on the basic dependency trees. Our approach combines language-specific BERT models, the UDify parser, neural sequence-to-sequence lemmatization and a graph transformation approach encoding the enhanced structure into a dependency tree. Our submission averaged 84.5% ELAS, ranking first in the shared task. We make all methods and resources developed for this study freely available under open licenses from https://turkunlp.org.

News articles such as sports game reports are often thought to closely follow the underlying game statistics, but in practice they contain a notable amount of background knowledge, interpretation, insight into the game, and quotes that are not present in the official statistics. This poses a challenge for automated data-to-text news generation with real-world news corpora as training data. We report on the development of a corpus of Finnish ice hockey news, edited to be suitable for training of end-to-end news generation methods, as well as demonstrate generation of text, which was judged by journalists to be relatively close to a viable product. The new dataset and system source code are available for research purposes.

The multilingual BERT model is trained on 104 languages and meant to serve as a universal language model and tool for encoding sentences. We explore how well the model performs on several languages across several tasks: a diagnostic classification probing the embeddings for a particular syntactic property, a cloze task testing the language modelling ability to fill in gaps in a sentence, and a natural language generation task testing for the ability to produce coherent text fitting a given context. We find that the currently available multilingual BERT model is clearly inferior to the monolingual counterparts, and cannot in many cases serve as a substitute for a well-trained monolingual model. We find that the English and German models perform well at generation, whereas the multilingual model is lacking, in particular, for Nordic languages. The code of the experiments in the paper is available at: https://github.com/TurkuNLP/bert-eval

We present the approach taken by the TurkuNLP group in the CRAFT Structural Annotation task, a shared task on dependency parsing. Our approach builds primarily on the Turku neural parser, a native dependency parser that ranked among the best in the recent CoNLL tasks on parsing Universal Dependencies. To adapt the parser to the biomedical domain, we considered and evaluated a number of approaches, including the generation of custom word embeddings, combination with other in-domain resources, and the incorporation of information from named entity recognition. We achieved a labeled attachment score of 89.7%, the best result among task participants.

We present our initial evaluation of a prototype system designed to assist nurses in assigning subject headings to nursing narratives – written in the context of documenting patient care in hospitals. Currently nurses may need to memorize several hundred subject headings from standardized nursing terminologies when structuring and assigning the right section/subject headings to their text. Our aim is to allow nurses to write in a narrative manner without having to plan and structure the text with respect to sections and subject headings, instead the system should assist with the assignment of subject headings and restructuring afterwards. We hypothesize that this could reduce the time and effort needed for nursing documentation in hospitals. A central component of the system is a text classification model based on a long short-term memory (LSTM) recurrent neural network architecture, trained on a large data set of nursing notes. A simple Web-based interface has been implemented for user interaction. To evaluate the system, three nurses write a set of artificial nursing shift notes in a fully unstructured narrative manner, without planning for or consider the use of sections and subject headings. These are then fed to the system which assigns subject headings to each sentence and then groups them into paragraphs. Manual evaluation is conducted by a group of nurses. The results show that about 70% of the sentences are assigned to correct subject headings. The nurses believe that such a system can be of great help in making nursing documentation in hospitals easier and less time consuming. Finally, various measures and approaches for improving the system are discussed.

In this paper, we focus on parsing rare and non-trivial constructions, in particular ellipsis. We report on several experiments in enrichment of training data for this specific construction, evaluated on five languages: Czech, English, Finnish, Russian and Slovak. These data enrichment methods draw upon self-training and tri-training, combined with a stratified sampling method mimicking the structural complexity of the original treebank. In addition, using these same methods, we also demonstrate small improvements over the CoNLL-17 parsing shared task winning system for four of the five languages, not only restricted to the elliptical constructions.

We evaluate two cross-lingual techniques for adding enhanced dependencies to existing treebanks in Universal Dependencies. We apply a rule-based system developed for English and a data-driven system trained on Finnish to Swedish and Italian. We find that both systems are accurate enough to bootstrap enhanced dependencies in existing UD treebanks. In the case of Italian, results are even on par with those of a prototype language-specific system.

Every year, the Conference on Computational Natural Language Learning (CoNLL) features a shared task, in which participants train and test their learning systems on the same data sets. In 2018, one of two tasks was devoted to learning dependency parsers for a large number of languages, in a real-world setting without any gold-standard annotation on test input. All test sets followed a unified annotation scheme, namely that of Universal Dependencies. This shared task constitutes a 2nd edition—the first one took place in 2017 (Zeman et al., 2017); the main metric from 2017 has been kept, allowing for easy comparison, also in 2018, and two new main metrics have been used. New datasets added to the Universal Dependencies collection between mid-2017 and the spring of 2018 have contributed to increased difficulty of the task this year. In this overview paper, we define the task and the updated evaluation methodology, describe data preparation, report and analyze the main results, and provide a brief categorization of the different approaches of the participating systems.

In this paper we describe the TurkuNLP entry at the CoNLL 2018 Shared Task on Multilingual Parsing from Raw Text to Universal Dependencies. Compared to the last year, this year the shared task includes two new main metrics to measure the morphological tagging and lemmatization accuracies in addition to syntactic trees. Basing our motivation into these new metrics, we developed an end-to-end parsing pipeline especially focusing on developing a novel and state-of-the-art component for lemmatization. Our system reached the highest aggregate ranking on three main metrics out of 26 teams by achieving 1st place on metric involving lemmatization, and 2nd on both morphological tagging and parsing.

The Conference on Computational Natural Language Learning (CoNLL) features a shared task, in which participants train and test their learning systems on the same data sets. In 2017, the task was devoted to learning dependency parsers for a large number of languages, in a real-world setting without any gold-standard annotation on input. All test sets followed a unified annotation scheme, namely that of Universal Dependencies. In this paper, we define the task and evaluation methodology, describe how the data sets were prepared, report and analyze the main results, and provide a brief categorization of the different approaches of the participating systems.

We present the TurkuNLP entry in the CoNLL 2017 Shared Task on Multilingual Parsing from Raw Text to Universal Dependencies. The system is based on the UDPipe parser with our focus being in exploring various techniques to pre-train the word embeddings used by the parser in order to improve its performance especially on languages with small training sets. The system ranked 11th among the 33 participants overall, being 8th on the small treebanks, 10th on the large treebanks, 12th on the parallel test sets, and 26th on the surprise languages.

Universal Dependencies (UD) is a project that seeks to develop cross-linguistically consistent treebank annotation for many languages. This tutorial gives an introduction to the UD framework and resources, from basic design principles to annotation guidelines and existing treebanks. We also discuss tools for developing and exploiting UD treebanks and survey applications of UD in NLP and linguistics.

We introduce an end-to-end system capable of named-entity detection, normalization and relation extraction for extracting information about bacteria and their habitats from biomedical literature. Our system is based on deep learning, CRF classifiers and vector space models. We train and evaluate the system on the BioNLP 2016 Shared Task Bacteria Biotope data. The official evaluation shows that the joint performance of our entity detection and relation extraction models outperforms the winning team of the Shared Task by 19pp on F1-score, establishing a new top score for the task. We also achieve state-of-the-art results in the normalization task. Our system is open source and freely available at https://github.com/TurkuNLP/BHE.

We study and compare two different approaches to the task of automatic assignment of predefined classes to clinical free-text narratives. In the first approach this is treated as a traditional mention-level named-entity recognition task, while the second approach treats it as a sentence-level multi-label classification task. Performance comparison across these two approaches is conducted in the form of sentence-level evaluation and state-of-the-art methods for both approaches are evaluated. The experiments are done on two data sets consisting of Finnish clinical text, manually annotated with respect to the topics pain and acute confusion. Our results suggest that the mention-level named-entity recognition approach outperforms sentence-level classification overall, but the latter approach still manages to achieve the best prediction scores on several annotation classes.

In this paper we present our system in the DiscoMT 2017 Shared Task on Crosslingual Pronoun Prediction. Our entry builds on our last year’s success, our system based on deep recurrent neural networks outperformed all the other systems with a clear margin. This year we investigate whether different pre-trained word embeddings can be used to improve the neural systems, and whether the recently published Gated Convolutions outperform the Gated Recurrent Units used last year.

Cross-linguistically consistent annotation is necessary for sound comparative evaluation and cross-lingual learning experiments. It is also useful for multilingual system development and comparative linguistic studies. Universal Dependencies is an open community effort to create cross-linguistically consistent treebank annotation for many languages within a dependency-based lexicalist framework. In this paper, we describe v1 of the universal guidelines, the underlying design principles, and the currently available treebanks for 33 languages.

The Persian Universal Dependency Treebank (Persian UD) is a recent effort of treebanking Persian with Universal Dependencies (UD), an ongoing project that designs unified and cross-linguistically valid grammatical representations including part-of-speech tags, morphological features, and dependency relations. The Persian UD is the converted version of the Uppsala Persian Dependency Treebank (UPDT) to the universal dependencies framework and consists of nearly 6,000 sentences and 152,871 word tokens with an average sentence length of 25 words. In addition to the universal dependencies syntactic annotation guidelines, the two treebanks differ in tokenization. All words containing unsegmented clitics (pronominal and copula clitics) annotated with complex labels in the UPDT have been separated from the clitics and appear with distinct labels in the Persian UD. The treebank has its original syntactic annotation scheme based on Stanford Typed Dependencies. In this paper, we present the approaches taken in the development of the Persian UD.

Revisiting the now de facto standard Stanford dependency representation, we propose an improved taxonomy to capture grammatical relations across languages, including morphologically rich ones. We suggest a two-layered taxonomy: a set of broadly attested universal grammatical relations, to which language-specific relations can be added. We emphasize the lexicalist stance of the Stanford Dependencies, which leads to a particular, partially new treatment of compounding, prepositions, and morphology. We show how existing dependency schemes for several languages map onto the universal taxonomy proposed here and close with consideration of practical implications of dependency representation choices for NLP applications, in particular parsing.

Joint morphological and syntactic analysis has been proposed as a way of improving parsing accuracy for richly inflected languages. Starting from a transition-based model for joint part-of-speech tagging and dependency parsing, we explore different ways of integrating morphological features into the model. We also investigate the use of rule-based morphological analyzers to provide hard or soft lexical constraints and the use of word clusters to tackle the sparsity of lexical features. Evaluation on five morphologically rich languages (Czech, Finnish, German, Hungarian, and Russian) shows consistent improvements in both morphological and syntactic accuracy for joint prediction over a pipeline model, with further improvements thanks to lexical constraints and word clusters. The final results improve the state of the art in dependency parsing for all languages.