108495 research outputs found
Sort by
Kroonilisi elustiilihaigusi ennetavad mittemedikamentoossed sekkumised: teaduskirjanduse ülevaade
No full textUAM-CSI at MultiGEC-2025: Parameter-efficient LLM Fine-tuning for Multilingual Grammatical Error Correction
Get PDFThis paper describes the solution of the UAMCSI team to the shared task on Multilingual Grammatical Error Correction (MultiGEC-2025), which is part of the workshop on Natural Language Processing for Computer-Assisted Language Learning (NLP4CALL). The shared task covers 12 languages: Czech, English, Estonian, German, Greek, Icelandic, Italian, Latvian, Russian, Slovene, Swedish and Ukrainian. The aim of the task is to correct errors in the provided texts. Our system is a google/gemma-2-9b-it model with 2 QLoRA adapters, one for the minimal-edit track and another for the fluency-edit track. Our solution achieves the best performance on the test sets on GLEU and F0.5 metrics for all languages and the best performance on the Scribendi Score metric except for the Greek language in the minimal-edit track
The 1st Workshop on Nordic-Baltic Responsible Evaluation and Alignment of Language Models
No full textAssessing the Similarity of Cross-Lingual Seq2Seq Sentence Embeddings Using Low-Resource Spectral Clustering
No full textIn this work, we study the cross-lingual distance of machine translations through alignment of seq2seq representations over small corpora. First, we use the M2M100 model to collect sentence-level representations of The Book of Revelation in several languages. We then perform unsupervised manifold alignment (spectral clustering) between these collections of embeddings. As verses between translations are not necessarily aligned, our procedure falls under the challenging, but more realistic non-correspondence regime. The cost function associated with each alignment is used to rank the relative (machine) similarity of one language to another. We then perform correspondent alignment over another cluster of languages, this time using FLORES+ parallel NLLB model embeddings. Our experiments demonstrate that the representations of closely-related languages group closely, and are cheap to align (requiring 1000 sentences) via our strategy.https://aclanthology.org/2025.resourceful-1.0
Multi-label Scandinavian Language Identification (SLIDE)
No full textIdentifying closely related languages at sentence level is difficult, in particular because it is often impossible to assign a sentence to a single language. In this paper, we focus on multi-label sentence-level Scandinavian language identification (LID) for Danish, Norwegian Bokmål, Norwegian Nynorsk, and Swedish. We present the Scandinavian Language Identification and Evaluation, SLIDE, a manually curated multi-label evaluation dataset and a suite of LID models with varying speed–accuracy tradeoffs. We demonstrate that the ability to identify multiple languages simultaneously is necessary for any accurate LID method, and present a novel approach to training such multi-label LID models.https://aclanthology.org/2025.resourceful-1.0
Mikroelupaigad elusatel puudel, nende kujunemine, püsimine ja seos metsa elurikkusega
No full textVäitekirja elektrooniline versioon ei sisalda publikatsiooneKäesolev doktoritöö uurib eluspuudega seotud mikroelupaikade (ingl. k. tree-related microhabitat) kujunemist, püsimist ja ökoloogilist tähtsust. Mikroelupaigad nagu puu-õõnsused, praod, tuuleluuad jm moodustised suurendavad metsaelupaikade mitmekesisust ja toetavad seeläbi metsaelustikku. Doktoritöö eesmärk on uurida, kuidas kujundavad mikroelupaigad eluspuude elupaigaväärtust erinevates oludes. Väliuuringud keskendusid kuusele ja haavale Eestis. Uuriti, kuidas puude ja puistute omadused mõjutavad mikroelupaikade esinemist ja mitmekesisust. Tööst selgub, et mikroelupaikade kujunemine on keeruline protsess, mida mõjutavad puu vanus, suurus, liik, elupaiga tingimused ja häiringud. Eesti majandusmetsades oli puudega seotud mikroelupaiku üldiselt vähe ja mõnda tüüpi välitööl ei leitudki. Kuigi sellised mikroelupaigad suurendavad elupaikade struktuurset keerukust, võib nende mõju bioloogilisele mitmekesisusele varieeruda, sõltuvalt puu vanusest ja uuritavast organismirühmast. Doktoritöö rõhutab ettevaatuspõhimõtte olulisust metsamajandamises, et edendada mikroelupaikade mitmekesisust eri puuliikidel ja vanuses. See hõlmab olemasolevate mikroelupaikadega puude säilitamist ja nende teket soodustavate looduslike protsesside edendamist maastiku tasandil, mis aitab kaasa metsaökosüsteemide pikaajalisele toimimisele, vastupanuvõimele ja bioloogilisele mitmekesisusele.This thesis investigates the development, persistence, and ecological significance of tree-related microhabitats (TreMs) in live trees – features like cavities, cracks, and loose bark that enhance forest habitat complexity and support a wide range of organisms. This thesis aims to gain a deeper understanding of how trees develop their habitat value through TreM dynamics. Focusing on Norway spruce and European aspen in Estonia, the research examines how tree characteristics and stand-level factors influence the occurrence and diversity of TreMs, revealing that TreM development is a complex process driven by factors such as tree age, size, species, habitat conditions, and disturbances. In managed forests in Estonia, there were generally few TreMs and some types were not found during fieldwork. While TreMs increase habitat complexity, their impact on biodiversity can vary depending on tree age and the organism group studied. This thesis highlights the importance of a precautionary approach to forest management to promote diverse TreMs across various tree species and ages. This includes retaining trees with existing TreMs and promoting natural processes that lead to TreM formation at the landscape level, ultimately contributing to forest ecosystems' long-term health, resilience, and biodiversity.https://www.ester.ee/record=b572857
Kunsti õpetamine lasteaias õpetajakeskseid õppemeetode kasutades
No full texthttps://www.ester.ee/record=b5732779*es
Thematic Categorization on Pineapple Production in Costa Rica: An Exploratory Analysis through Topic Modeling
No full textCosta Rica is one of the largest producers and exporters of pineapple in the world. This status has encouraged multinational companies to use plantations in this Central American country for experimentation and the cultivation of new varieties, such as the Pinkglow pineapple. However, pineapple monoculture has significant socio-environmental impacts on the regions where it is cultivated.In this exploratory study, we aimed to analyze how pineapple production is portrayed on the Internet. To achieve this, we collected a corpus of texts in Spanish and English from online sources in two phases: using the BootCat tool and manual search on newspaper websites. The Hierarchical Dirichlet Process (HDP) topic model was then applied to identify dominant topics within the corpus. These topics were subsequently classified into thematic categories, and the texts were categorized accordingly. The findings indicate that environmental issues related to pineapple cultivation are underrepresented on the Internet, particularly in comparison to the extensive focus on topics related to pineapple production and marketing
Plant phylogeny, traits and fungal community composition asdrivers of plant–soil feedbacks
No full textPlant–soil feedbacks (PSFs) are key drivers of plant community dynamics. However, our understanding of the factors moderating PSFs remains limited. We examined how plant phylogenetic relatedness and functional traits determine PSFs via their influence on rhizosphere fungal communities, especially arbuscular mycorrhizal fungi (AMF) and fungal pathotrophs.
We conducted a glasshouse PSF experiment using 21 temperate grassland plant species, where each focal species was exposed to soils conditioned by heterospecific plants of increasing phylogenetic dissimilarity. We tested whether phylogenetic distance between plant species, functional traits or the degree to which species associate with AMF or fungal pathotrophs, explained the magnitude and direction of PSF responses.
None of the measured plant traits explained PSFs, although the relative abundance of AMF was weakly and positively related to PSFs. Across all plant species, phylogenetic relatedness did not explain PSFs. However, species-specific effects of phylogenetic relatedness on the outcome of PSFs were detected. In particular, significant relationships with phylogenetic relatedness were observed only for species characterised by the highest rhizosphere relative abundance of AMF or fungal pathotrophs. For Centaurea nigra and Vicia cracca (both high AMF abundance) and Anthoxanthum odoratum (high pathotroph abundance), we found that PSF became more positive with increased phylogenetic distance between focal and conditioning species, showing a shift towards increased performance in conspecific than heterospecific soils. Meanwhile, as phylogenetic dissimilarity between Poa trivialis (high pathotroph abundance) and the soil conditioning species increased, more negative PSFs were observed, indicating improved performance in soils conditioned by increasingly distant heterospecific species.
Synthesis. Our results suggest that plant traits and phylogenetic relatedness are poor predictors of plant–soil feedbacks (PSFs) among temperate grassland plant species. Hence, despite known effects of these factors in shaping rhizosphere microbial communities, the way plant species respond to these microbial communities is not related to the same characteristics. The occurrence of significant relationships between phylogenetic distance and PSFs in species with high relative abundances of mycorrhizal or pathogenic fungi suggests that the tendency to accumulate fungal mutualists or pathotrophs may be an important moderator of the relationship between plant phylogenetic relatedness and the magnitude and direction of PSFs