Prispevek se ukvarja z vprašanjem, kako pri pouku slovenščine kot tujega jezika obravnavati klasična literarna besedila. Izpostavi specifike tovrstnih besedil, zaradi katerih mora njihovo ...vključevanje v pouk potekati postopno in načrtovano. Analizira naloge, ki so v obstoječih učbenikih slovenščine kot drugega in tujega jezika pripravljene ob besedilih Ivana Cankarja. Ob odlomku iz Cankarjevega romana Na klancu pa predlaga dejavnosti, ki omogočajo tako razvijanje sporazumevalne jezikovne zmožnosti kot tudi estetsko doživljanje literarnega dela.
Poročilo o 34. evropski poletni šoli logike, jezika in informatike (European Summer School of Logic, Language and Information (ESSLLI)), ki je potekala med 31. julijem in 11. avgustom 2023 na ...Fakulteti za računalništvo in informatiko v Ljubljani.
Coreference resolution systems aim to recognize and cluster together mentions of the same underlying entity. While there exist large amounts of research on broadly spoken languages such as English ...and Chinese, research on coreference in other languages is comparably scarce. In this work we first present SentiCoref 1.0 - a coreference resolution dataset for Slovene language that is comparable to English-based corpora. Further, we conduct a series of analyses using various complex models that range from simple linear models to current state-of-the-art deep neural coreference approaches leveraging pre-trained contextual embeddings. Apart from SentiCoref, we evaluate models also on a smaller coref149 Slovene dataset to justify the creation of a new corpus. We investigate robustness of the models using cross-domain data and data augmentations. Models using contextual embeddings achieve the best results - up to 0.92 average F1 score for the SentiCoref dataset. Cross-domain experiments indicate that SentiCoref allows the models to learn more general patterns, which enables them to outperform models, learned on coref149 only.
Abstract
Deep learning approaches are superior in natural language processing due to their ability to extract informative features and patterns from languages. The two most successful neural ...architectures are LSTM and transformers, used in large pretrained language models such as BERT. While cross-lingual approaches are on the rise, most current natural language processing techniques are designed and applied to English, and less-resourced languages are lagging behind. In morphologically rich languages, information is conveyed through morphology, for example, through affixes modifying stems of words. The existing neural approaches do not explicitly use the information on word morphology. We analyse the effect of adding morphological features to LSTM and BERT models. As a testbed, we use three tasks available in many less-resourced languages: named entity recognition (NER), dependency parsing (DP) and comment filtering (CF). We construct baselines involving LSTM and BERT models, which we adjust by adding additional input in the form of part of speech (POS) tags and universal features. We compare the models across several languages from different language families. Our results suggest that adding morphological features has mixed effects depending on the quality of features and the task. The features improve the performance of LSTM-based models on the NER and DP tasks, while they do not benefit the performance on the CF task. For BERT-based models, the added morphological features only improve the performance on DP when they are of high quality (i.e., manually checked) while not showing any practical improvement when they are predicted. Even for high-quality features, the improvements are less pronounced in language-specific BERT variants compared to massively multilingual BERT models. As in NER and CF datasets manually checked features are not available, we only experiment with predicted features and find that they do not cause any practical improvement in performance.
This volume offers a thematic exploration of the migrant artist's experience in Europe and its colonies from the early modern period through to the Industrial Revolution. The influence of the ...transient artist, both on their adoptive country as well as their own oeuvre and native culture, is considered through a collection of essays arranged according to geographic location. The contributions here examine the impetuses behind artistic migrations and the status of the foreign artist at home and abroad through the patterns of patronage, contemporary responses to their work and the preservation of their artistic legacy in domestic and foreign settings. Objects and sites from across the visual arts are considered as evidence of the migrant artist's experience; talismans of cultural exchange that yielded hybrid artistic styles and disseminated foreign tastes and workshop practices across the globe.
Paraphrasing is a useful natural language processing task that can contribute to more diverse generated or translated texts. Natural language inference (NLI) and paraphrasing share some similarities ...and can benefit from a joint approach. We propose a novel methodology for the extraction of paraphrasing datasets from NLI datasets and cleaning existing paraphrasing datasets. Our approach is based on bidirectional entailment; namely, if two sentences can be mutually entailed, they are paraphrases. We evaluate our approach using several large pretrained transformer language models in the monolingual and cross-lingual setting. The results show high quality of extracted paraphrasing datasets and surprisingly high noise levels in two existing paraphrasing datasets.
Deep learning approaches are superior in NLP due to their ability to extract informative features and patterns from languages. The two most successful neural architectures are LSTM and transformers, ...used in large pretrained language models such as BERT. While cross-lingual approaches are on the rise, most current NLP techniques are designed and applied to English, and less-resourced languages are lagging behind. In morphologically rich languages, information is conveyed through morphology, e.g., through affixes modifying stems of words. Existing neural approaches do not explicitly use the information on word morphology. We analyse the effect of adding morphological features to LSTM and BERT models. As a testbed, we use three tasks available in many less-resourced languages: named entity recognition (NER), dependency parsing (DP), and comment filtering (CF). We construct baselines involving LSTM and BERT models, which we adjust by adding additional input in the form of part of speech (POS) tags and universal features. We compare models across several languages from different language families. Our results suggest that adding morphological features has mixed effects depending on the quality of features and the task. The features improve the performance of LSTM-based models on the NER and DP tasks, while they do not benefit the performance on the CF task. For BERT-based models, the morphological features only improve the performance on DP when they are of high quality while not showing practical improvement when they are predicted. Even for high-quality features, the improvements are less pronounced in language-specific BERT variants compared to massively multilingual BERT models. As in NER and CF datasets manually checked features are not available, we only experiment with predicted features and find that they do not cause any practical improvement in performance.
Code-mixed discourse combines multiple languages in a single text. It is commonly used in informal discourse in countries with several official languages, but also in many other countries in ...combination with English or neighboring languages. As recently large language models have dominated most natural language processing tasks, we investigated their performance in code-mixed settings for relevant tasks. We first created four new bilingual pre-trained masked language models for English-Hindi and English-Slovene languages, specifically aimed to support informal language. Then we performed an evaluation of monolingual, bilingual, few-lingual, and massively multilingual models on several languages, using two tasks that frequently contain code-mixed text, in particular, sentiment analysis and offensive language detection in social media texts. The results show that the most successful classifiers are fine-tuned bilingual models and multilingual models, specialized for social media texts, followed by non-specialized massively multilingual and monolingual models, while huge generative models are not competitive. For our affective problems, the models mostly perform slightly better on code-mixed data compared to non-code-mixed data.