• Digital Humanities and Cultural Heritage
• 2012-2021close
• Publications close
• Research data close
• Other research products close
• EE close
• DSpace at Tartu University Library close
• Digital Humanities and Cultural Her... close

Siinne doktoritöö käsitleb rahvuskuuluvuse tähendust inimeste ja riigi vahelistes suhetes 20. sajandi esimesel poolel, kasutades näidisvalimina 1941. aastal Eestist Saksamaale ümber asunud inimeste andmeid. Uurimuse eesmärk on selgitada, kas ja kuidas riiklik rahvuspoliitika mõjutas inimeste käitumist ja vastupidi – mil määral mõjutasid inimeste valikud riikide tegevust. Valimisse hõlmatud inimestest suurem osa puutus uuritaval perioodil kokku nelja riigi – Venemaa keisririigi, Eesti Vabariigi, Nõukogude Liidu ja Saksamaa – rahvuspoliitikaga. Nende põhimõtteliselt erineva rahvuspoliitikaga riikide käsitlemine võimaldab tuvastada rahvuse ühiskondliku tähenduse universaalseid jooni. Eesmärgi täitmiseks analüüsiti inimeste eluloolisi andmeid, seadusandlust ja rahvuspoliitikat puudutavat dokumentatsiooni ning ühiskondlikke olusid puudutavaid allikaid. Doktoritööst selgus, et üksikisikute rahvuskuuluvuse formaalne määramine osutus 20. sajandi esimesel poolel keeruliseks nii riikide kui ka inimeste jaoks, sest puudusid selged kriteeriumid viimaste rahvuspõhiseks kategoriseerimiseks, ühtsed arusaamad rahvuste olemusest ning takistav tegur oli ka mõlema poole pragmaatiliste kaalutluste domineerimine rahvuse määramisel. Mida suurem oli rahvuskuuluvuse formaalne tähtsus, seda enam olid üksikisikute valikud tingitud olukorrast ning märksa vähem nende harjumuspärasest keele- ja kultuurikeskkonnast. Seetõttu ei õnnestunud riikidel peaaegu kunagi soovitud määral rakendada essentsialistlikke ja objektiivsusele pretendeerinud rahvuse määramise kriteeriume, isegi kui see oli eesmärk ja taolisi põhimõtteid seadustega kehtestada püüti. Samuti ei olnud riigid objektiivsuse rakendamisel põhimõttelised ja järjekindlad, vaid kohandasid seisukohti vastavalt pragmaatilistele huvidele ning inimeste käitumisele. Üksikisikute valikutel oli riiklikele poliitikatele märkimisväärne mõju, ehkki arvuliselt oli muutliku ja ebamäärase rahvuskuuluvusega inimesi ühiskonnas marginaalselt. This dissertation focuses on the meaning of ethnicity in the relations between individuals and states in the first half of the 20th century using the biographical data of people who resettled from Estonia to Germany in 1941 as a sample. The aim of this study is to explain whether and how ethnopolitics of the states influenced the behaviour of individuals and vice versa – to what extent individual choices influenced state policy. Most people in the sample were affected by the ethnopolitics of four states: Russian Empire, Estonian Republic, Soviet Union, and Germany. Analysing the states with fundamentally different ethnopolitics enables to identify universal characteristics of the role of ethnicity in the society. To fulfil the aim, biographical data, legislation, and documentation about ethnopolitics as well as sources about the social conditions were analysed. The dissertation shows that fixing individuals’ ethnicity turned out to be difficult for both parties – the states and individuals – since there were no clear criteria for ethnic categorisation and no common understanding about the nature of ethnicity. Pragmatic considerations of both parties also turned out to be an obstacle. The more formal importance ethnicity had, the more individuals made choices dependent on the situation and not so much on their customary lingual and cultural environment. Therefore, the states almost always failed to implement essentialist and objective criteria in fixing individuals’ ethnicity in a desired manner, even if such criteria were their aim and established by law. Furthermore, the authorities of the states were neither principled nor consistent in applying objective criteria. The states had to adapt their stances to their own pragmatic interests and according to the behaviour of individuals instead. Thus, individual choices had a remarkable impact on ethnopolitics, although people with dynamic and vague ethnic identity formed a marginal group in the society. https://www.ester.ee/record=b5450809

Advanced NLP models require huge amounts of data from various domains to produce high-quality representations. It is useful then for a few large public and private organizations to join their corpora during training. However, factors such as legislation and user emphasis on data privacy may prevent centralized orchestration and data sharing among these organizations. Therefore, for this specific scenario, we investigate how gossip learning, a massively-parallel, data-private, decentralized protocol, compares to a shared-dataset solution. We find that the application of Word2Vec in a gossip learning framework is viable. Without any tuning, the results are comparable to a traditional centralized setting, with a reduction in ground-truth similarity scores as low as 4.3%. Furthermore, the results are up to 54.8% better than independent local training. QC 20210423

In this paper, using Estonian Community Innovation Survey data, we study the role of R&D, capital accumulation, and innovation output on pro- ductivity for entrants and incumbents. We find that the impact of R&D invest- ment on labour productivity is larger for the entrants compared to the incum- bents. Entrants are found to be more productive and more heterogeneous in their total factor productivity (TFP) than the incumbents. Moreover, entrants who innovate are on average, in terms of TFP, 25% more productive than the entrants who do not, while the corresponding figure for the incumbents is 7%. In addition, it is mostly the incumbents who benefit from within-industry knowledge that is produced outside their own firm. Finally, for both entrants and incumbents, em- bodied technological change through capital accumulation is found to be more effective in generating productivity growth than R&D expenditure.

For stock market predictions, the essence of the problem is usually predicting the magnitude and direction of the stock price movement as accurately as possible. There are different approaches (e.g., econometrics and machine learning) for predicting stock returns. However, it is non-trivial to find an approach which works the best. In this paper, we make a thorough analysis of the predictive accuracy of different machine learning and econometric approaches for predicting the returns and volatilities on the OMX Baltic Benchmark price index, which is a relatively less researched stock market. Our results show that the machine learning methods, namely the support vector regression and k-nearest neighbours, predict the returns better than autoregressive moving average models for most of the metrics, while for the other approaches, the results were not conclusive. Our analysis also highlighted that training and testing sample size plays an important role on the outcome of machine learning approaches.

This paper investigates the productivity effects for domestic suppliers from joining and exiting the value chains of multinational enterprises (MNEs). The vast majority of prior literature has relied on sector-level input-output tables in estimating the effects of vertical linkages of FDI. Instead, our econometric analysis of the creation and destruction of backward linkages of MNEs is based on information on firm-to-firm transactions recorded in the valued added tax declarations data. Treatment analysis based on propensity score matching and panel data from Estonia suggests that starting to supply multinationals initially boosts the value added per employee of domestic firms, including effects on the scale of production and the capital-labour ratio. These first linkages to MNEs do not affect the total factor productivity (TFP) of domestic firms, suggesting that TFP effects take time to materialise. We further find that there are limits to the wider diffusion of the effects of linkages to MNEs. We find no significant positive effects on the second-tier suppliers: the positive effects are limited to the first-tier suppliers with direct links to MNEs. One novel result is the evidence that the productivity of suppliers does not fall, on average, after decreasing or ending supplier relationships with MNE customers.

Can a country with a constant state of emergency in its capital – as was the case in Estonia in 1918–1934 – be considered a democracy? Or a country where political violence is used against members of a parliamentary committee during a committee meeting – as was the case in Finland in 1930? Why did the Estonian Social Democrats support the 1934 coup and why did the Finnish Swedes defend the Finnish Communists? Why could Estonia not find any allies except Latvia in the 1920s and 1930s, and why was Finland left without allies in the Winter War? Why did Estonia and Finland not behave in the same way in the autumn of 1939, and why did the United Kingdom, Canada, Australia, and New Zealand declare war on Finland in 1941? The answers to these questions lie in the relationships between democracy and security, which is exactly the topic of my dissertation. Based on the works of different philosophers, it is possible to show that democracy and security are interdependent. The relationship between democracy and security can be reduced to the classic dilemmas between freedom and security, freedom and sovereignty, and sovereignty and security. Although it is not possible to solve these dilemmas definitively, the solution can be improved indefinitely. I argue that the way in which democratic societies resolve these dilemmas does not depend on democratic culture or on the wisdom inherent in people, but on the perceptions of the majorities of democratic societies, which in turn depend primarily on historical experience. The theoretical explanation proposed in my dissertation helps to understand why Estonia’s and Finland’s quests for security failed in 1918–1948 and why Estonia and Finland made different choices at the turning points of their history. The behaviour of Estonia’s leaders during the events of 1939 and 1940 also becomes understandable. Hopefully, my dissertation will contribute to a fair judgement of the activities of Konstantin Päts and Johan Laidoner, and to reconciling Estonian society on the question of whether they were heroes or villains. However, the most important aim is to contribute to an understanding of just how fragile democracy and security can be, how they depend on each other and, in particular, how they depend on historical experience. Kas saab pidada demokraatlikuks riiki, mille pealinnas kehtib pidev erakorraline seisukord – nagu oli Eesti aastatel 1918–1934? Või riiki, kus parlamendikomisjoni liikmete vastu kasutatakse poliitilist vägivalda komisjoni istungi toimumise ajal – nagu oli Soome aastal 1930? Miks Eesti sotsiaaldemokraadid toetasid 1934. aasta riigipööret ja miks soomerootslased kaitsesid Soome kommuniste? Miks ei õnnestunud 1920-ndate ja 1930-ndate Eestil leida ühtegi liitlast peale Läti ja miks jäi Soome talvesõjas ilma liitlasteta? Miks Eesti ja Soome 1939. aasta sügisel ühtemoodi ei käitunud ja miks Ühendkuningriik, Kanada, Austraalia ja Uus-Meremaa 1941. aastal Soomele sõja kuulutasid? Vastus peitub demokraatia ja julgeoleku vahelistes seostes, mida minu väitekiri uuribki. Erinevate filosoofide töödele tuginedes on võimalik näidata, et demokraatia ja julgeolek sõltuvad vastastikku üksteisest. Demokraatia ja julgeoleku vaheline seos on taandatav klassikalistele dilemmadele vabaduse ja julgeoleku, vabaduse ja suveräänsuse ja suveräänsuse ja julgeoleku vahel. Kuigi dilemmasid ei ole võimalik lõplikult lahendada, on lahendust võimalik lõputult täiustada. Ma väidan, et see, kuidas demokraatlikud ühiskonnad neid dilemmasid lahendavad, ei sõltu mitte demokraatlikust kultuurist või rahvale omasest tarkusest, vaid demokraatlike ühiskondade enamuste arusaamadest, mis omakorda sõltuvad eelkõige ajaloolisest kogemusest. Minu väitekirjas välja pakutud teoreetiline selgitus aitab mõista, miks Eesti ja Soome julgeolekuotsingud aastatel 1918–1948 läbi kukkusid ja miks Eesti ja Soome oma ajaloo pöördepunktides erinevaid valikuid tegid. Mõistetavaks muutub ka Eesti riigijuhtide käitumine 1939. ja 1940. aasta sündmuste ajal. Loodetavasti aitab minu väitekiri kaasa õiglase hinnangu andmisele Konstantin Pätsi ja Johan Laidoneri tegevusele ja Eesti ühiskonna lepitamisele küsimuses, kas tegemist oli kangelaste või kurjategijatega. Kõige tähtsam on siiski aidata kaasa mõistmisele, kuivõrd haprad võivad olla demokraatia ja julgeolek, kuidas nad vastastikku üksteisest sõltuvad ja eriti – kuidas nad sõltuvad ajaloolisest kogemusest. https://www.ester.ee/record=b5391015

Deep neural language models such as BERT have enabled substantial recent advances in many natural language processing tasks. Due to the effort and computational cost involved in their pre-training, language-specific 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 the state-of-the-art UDify parser on Universal Dependencies data, contrasting performance with results using the multilingual BERT model. We find 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. We also present preliminary results as first steps toward an understanding of the conditions under which language-specific models are most beneficial. All of the methods and models introduced in this work are available under open licenses from https://github.com/turkunlp/wikibert. Comment: 7 pages, 1 figure

This thesis is based on three main objectives: at first, the implementation of RNMT+ archi-tecture with Relational-RNN model. This is an interaction between this architecture and the RNN model. Secondly, train three different translation models based on RNMT+, Trans-former, and sequence to sequence architectures. Previously, we have witnessed the perfor-mance comparison among RNMT+ with LSTM, Transformer, seq2seq, etc. Finally, evalu-ate the translation model based on training data. When implementing RNMT+, the core idea was to use a newer type of Recurrent Neural Network (RNN) instead of a widely used LSTM or GRU. Besides this, we evaluate the RNMT+ model with other models based on state-of-the-art Transformer and Sequence to Sequence with attention architectures. This evaluation (BLEU) shows that neural machine translation is domain-dependent, and translation based on the Transformer model performs better than the other two in OpenSubtitle v2018 domain while RNMT+ model performs better compared to other two in a cross-domain evaluation. Additionally, we compare all the above-mentioned architectures based on their correspond-ing encoder-decoder layers, attention mechanism and other available neural machine translation and statistical machine translation architectures. In estonian: See lõputöö põhineb kolmel põhieesmärgil: alguses RNMT + arhitektuuri rakendamine Relatsioon-RNN-mudeli abil. See on interaktsioon selle arhitektuuri ja RNN-mudeli vahel. Teiseks, koolitage kolme erinevat tõlkemudelit, mis põhinevad RNMT +, Trafo ja järjestusearhitektuuridel. Varem oleme olnud tunnistajaks RNMT + jõudluse võrdlusele LSTM, Transformeri, seq2seq jne abil. Lõpuks hinnake tõlkemudelit koolitusandmete põhjal. RNMT + rakendamisel oli peamine idee kasutada laialdaselt kasutatava LSTM või GRU asemel uuemat tüüpi korduvat närvivõrku (RNN). Lisaks hindame RNMT + mudelit koos teiste mudelitega, mis põhinevad tipptehnoloogial Transformer ja Sequence to Sequence koos tähelepanu arhitektuuridega. See hinnang (BLEU) näitab, et neuraalne masintõlge on domeenist sõltuv ja muunduril Transformer põhinev tõlge toimib paremini kui ülejäänud kaks OpenSubtitle v2018 domeenis, samal ajal kui RNMT + mudel toimib paremini kui ülejäänud kaks domeenidevahelist hindamist. Lisaks võrdleme kõiki ülalnimetatud arhitektuure nende vastavate kodeerija-dekoodri kihtide, tähelepanu mehhanismi ja muude saadaolevate närvi masintõlke ning statistiliste masintõlke arhitektuuride põhjal.

http://tartu.ester.ee/record=b2658242~S1*est