Logo: to the web site of Uppsala University

Representational Similarity Analysis (RSA) is a technique developed by neuroscientists for comparing activity patterns of different measurement modalities (e.g., fMRI, electrophysiology, behavior). As a framework, RSA has several advantages over existing approaches to interpretation of language encoders based on probing or diagnostic classification: namely, it does not require large training samples, is not prone to overfitting, and it enables a more transparent comparison between the representational geometries of different models and modalities. We demonstrate the utility of RSA by establishing a previously unknown correspondence between widely-employed pre-trained language encoders and human processing difficulty via eye-tracking data, showcasing its potential in the interpretability toolbox for neural models.

Recent studies have shown that language models pretrained and/or fine-tuned on randomly permuted sentences exhibit competitive performance on GLUE, putting into question the importance of word order information. Somewhat counter-intuitively, some of these studies also report that position embeddings appear to be crucial for models' good performance with shuffled text. We probe these language models for word order information and investigate what position embeddings learned from shuffled text encode, showing that these models retain information pertaining to the original, naturalistic word order. We show this is in part due to a subtlety in how shuffling is implemented in previous work - before rather than after subword segmentation. Surprisingly, we find even Language models trained on text shuffled after subword segmentation retain some semblance of information about word order because of the statistical dependencies between sentence length and unigram probabilities. Finally, we show that beyond GLUE, a variety of language understanding tasks do require word order information, often to an extent that cannot be learned through fine-tuning.

I mitten av 2017 uppstod en diplomatisk kris mellan Qatar och fyra grannländer, nämligen Saudiarabien, Förenade Arabemiraten, Bahrain och Egypten. Som följd av krisen avbröt dessa länder sina diplomatiska förbindelser med Qatar.Syftet med denna uppsats är att undersöka innehållet i nyhetsartiklar om den inträffade diplomatiska krisen från både Al-Jazeera och Al-Arabiya samt att försöka kartlägga hur rapporteringen skilde sig åt. De primära frågeställningarna i denna undersökning kommer att behandla vilka normer de två tv-kanalerna har använt sig av för att producera sina respektive nyhetsstoff samt vilka tillvägagångssätt som har tillämpats för att publicera eller rapportera till respektive publik. Vidare kommer undersökningen att försöka kartlägga, analysera och skildra andra aspekter av värde såsom objektivitetsnivån, hur finansieringen påverkade de två nyhetskanalernas nyhetskvalitet samt deras spridning. Undersökningen kommer att granska om nyhetsartiklarna i båda kanalerna har gett en nyanserad bild eller bara rapporterat en sida av händelsen. Samtliga frågeställningar har sammanfattats i två huvudpunkter. Den första punkten syftar till att klargöra vilka språkliga skillnader som förekommer i rapporteringen om Qatarkrisen i de två satellit-tv-kanalerna Al-Jazeeras och Al-Arabiyas nyhetsartiklar. Den andra punkten avser att undersöka om Al-Jazeera och Al-Arabiya har varit neutrala eller påverkade av sina respektive regeringars politiska inställning i rapporteringen om Qatarkrisen.Jag har analyserat fem artiklar från varje kanal för att se vilken information som förmedlas. Som jag kommer att förtydliga, finns det tydliga tendenser som visar att rapporteringen av vissa händelseförlopp i Qatarkrisen skiljer sig avsevärt åt, vilket eskalerade krisen. Utifrån innehållsanalys och temaanalys med kodning presenterar jag en jämförelse av hur språket och ordvalet skiljer sig i nyhetsrapporteringen. Både innehållsanalys och tematisk analys har samma mål, nämligen att analysera material genom att dela upp texten i små innehållsenheter och sedan analysera texterna (Vaismoradi et al, 2013, ss. 400–402). Resultatet av denna jämförelse visar tydliga skillnader i ordval och parafraser av samma nyheter mellan de båda kanalerna. Till exempel rapporterade Al-Jazeera i den första nyheten att Qatars nyhetsbyrå hade hackats och att falska uttalanden av emiren hade publicerats. Al-Arabiya undvek dock att nämna att nyhetsbyrån hade hackats och rapporterade istället att alla uttalanden av emiren var falska.Denna studie baseras på tidigare forskning av forskarna Leon Brakho och Kristina Riegert, som har skrivit flera böcker om arabiska massmedier. Dessutom har jag använt tre olika teorier för att besvara frågeställningarna, nämligen normativ teori, auktoritär teori och gestaltningsteori.Denna studie baseras på tidigare forskning av forskarna Leon Brakho och Kristina Riegert, som har skrivit flera böcker om arabiska massmedier. Dessutom har jag använt tre olika teorier för att besvara frågeställningarna, nämligen normativ teori, auktoritär teori och gestaltningsteori.Genom jämförelsen mellan nyhetsartiklarna åskådliggörs att kanalerna har använt olika termer och ord om samma händelse. Samtidigt förekommer mer motstridig terminologi i rapporteringen mellan de båda nyhetskanalerna. Det var anmärkningsvärt att se hur uppbyggnaden av varje nyhetsartikel, samt hur varje kanal skildrar eller rapporterar om en viss händelse, skilde sig åt.

This is a comprehensive study on modality in one of the largest Turkic languages, Kazakh, as it is spoken in China. Kazakh is the official language of the Republic of Kazakhstan and is furthermore spoken by about one and a half million people in China in the Xinjiang Uyghur Autonomous Region and in Aksai Kazakh Autonomous County in Gansu Province.The method employed is empirical, i.e. data-oriented. The modal expressions in Kazakh are analyzed in a theoretical framework essentially based on the works of Lars Johanson. The framework defines semantic notions of modality from a functional and typological perspective. The modal volition, deontic evaluation, and epistemic evaluation express attitudes towards the propositional content and are conveyed in Kazakh by grammaticalized moods, particles and lexical devices. All these categories are treated in detail, and ample examples of their different usages are provided with interlinear annotation. The Kazakh expressions are compared with corresponding ones used in other Turkic languages. Contact influences of Uyghur and Chinese are also dealt with.The data used in this study include texts recorded by the author in 20102012, mostly in the northern regions of Xinjiang, as well as written texts published in Kazakhstan and China. The written texts represent different genres: fiction, non-fiction, poetry and texts published on the Internet. Moreover, examples have been elicited from native speakers of Kazakh and Uyghur. The Appendix contains nine texts recorded by the author in the Kazakh-speaking regions of Xinjiang, China. These texts illustrate the use of many of the items treated in the study.

This is a comprehensive study on expressions of modality in one of the largest Turkic languages, Kazakh, as it is spoken in China. Kazakh is the official language of the Republic of Kazakhstan and is furthermore spoken by about one and a half million people in China in the Xinjiang Uyghur Autonomous Region and in Aksai Kazakh Autonomous County in Gansu Province.The method employed is empirical, i.e. data-oriented. The modal expressions in Kazakh are analyzed in a theoretical framework essentially based on the works of Lars Johanson. The framework defines semantic notions of modality from a functional and typological perspective. The modal volition, deontic evaluation, and epistemic evaluation express attitudes towards the propositional content and are conveyed in Kazakh by grammaticalized moods, particles and lexical devices. All these categories are treated in detail, and ample examples of their different usages are provided with interlinear annotation. The Kazakh expressions are compared with corresponding ones used in other Turkic languages. Contact influences of Uyghur and Chinese are also dealt with.The data used in this study include texts recorded by the author in 20102012, mostly in the northern regions of Xinjiang, as well as written texts published in Kazakhstan and China. The written texts represent different genres: fiction, non-fiction, poetry and texts published on the Internet. Moreover, examples have been elicited from native speakers of Kazakh and Uyghur.

The Appendix contains nine texts recorded by the author in the Kazakh-speaking regions of Xinjiang, China. These texts illustrate the use of many of the items treated in the study.

This chapter introduces the rest of the book by rapidly surveying the names of the Uralic languages and their speakers’ numbers, as well as discussing the relative size and age of the family and its nine branches. It outlines the prehistory of these branches and then goes on to sketch in outline the phonological and lexicogrammatical features that the editors think most interesting, problematic, suggestive, and instructive. The reader should come away from reading this chapter with an understanding of the four kinds of information contained in the rest of the book. These are (i) how and which speech sounds are used (vowels and vowel harmony, stress, consonants, and consonant gradation); (ii) how nouns and verbs change their shapes in sentences—case, number, and predestination (in nouns) and tense, mood, and object indexing (in verbs); (iii) how words are put together to make clauses and complex sentences; and (iv) how new words are made (derivation and compounding). The chapter also introduces the more important theoretical conventions that the editors have found most useful in characterizing Uralic languages. To make the chapter accessible to as wide as possible a readership, most of the terminology used is cast in a basic linguistic theory framework. 

The Uralic Languages, second edition, is a reference book which brings together detailed discussions of the historical development and specialized linguistic structures and features of the languages in the Uralic family.

The Uralic languages are spoken today in a vast geographical area stretching from Dalarna County in Sweden to Dudinka, Taimyr, Russia. There are currently approximately 50 languages in the group, the largest one among them being the state languages Finnish, Estonian, and Hungarian; other Uralic languages covered in the book are South Saami, Skolt Saami, Võro, Moksha Mordvin, Mari, Udmurt, Zyrian Komi, Mansi, Khanty, Nganasan, Forest and Tundra Enets, Nenets, and Selkup. The book also contains a chapter on Finnic languages, the reconstruction of Uralic, the history of Uralic studies, connections of Uralic to other language families, and language names, demographics, and degrees of endangerment. This second and thoroughly revised edition updates and augments the authoritative accounts of the first edition and reflects recent and ongoing developments in linguistics and the languages themselves, as well as our further enhanced understanding of the relations and patterns of influence between them. Each chapter combines modern linguistic analysis and documentary linguistics; a relatively uniform structure allows for easy typological comparison between the individual languages.

Written by an international team of experts, The Uralic Languages will be invaluable to students and researchers within linguistics, folklore, and Siberian studies.

This thesis focuses on the lived multilingualism of a group of young Russian-German adults who as children migrated together with their families from post-Soviet states to the Federal Republic of Germany during the 1990s. Today these adults live in a ‘Plattenbau’ housing estate in a small town in one of the new federal states of Germany. The large pre-fabricated concrete-slab system-built housing estates that were built during the GDR-era are today generally considered as deprived areas due a combination of decreasing population and high unemployment.

This thesis shows how young Russian-Germans create a multilingual community of practice and use various aspects of language and non-linguistic resources for identity construction. The data analysed in this thesis comes from ethnographic studies conducted during three phases of fieldwork between Spring 2011 and Spring 2012. The data was collected at a youth centre where the group of young Russian-German adults regularly met.

Combining intensive participant observation, field notes, photos, and narrative interviews the thesis is a mixed-method investigation. Underpinning the analysis of the research data are theoretical models of the relationship between language, identity, and space. Methodologically this study combines linguistic ethnography, narrative analysis, and membership categorization analysis.

The thesis argues that an ethnographical-narrative approach is a powerful tool that is able to highlight the role of language(s) and non-linguistic resources for identity construction in social spaces, illustrates how young Russian-Germans construct a web of multilingual identities by using social categories to position themselves and others, and shows how the lived multilingualism of  young Russian-German adults influences all aspects of their social lives. For example, the thesis shows the maintenance of Russian as a heritage language within Russian-German families, yet and an avoidance of visible signs of the Russian-German heritage in public spaces.

In the Swedish school system, mother tongue education has been available, under certain regulations, as an elective subject, since the Home Language Reform of 1977 (Reath Warren, 2017). Since its introduction, the subject has been the subject of debate and contestation (Salö et al., 2018) but nonetheless, researchers have found that mother tongue instruction offers opportunities for development of literacies in the mother tongue and other languages as well as supporting learning in general (see Utbildningsdepartementet, 2019, for an overview of research). Among the structural challenges the subject faces, is that there are no specific qualifications required to become a mother tongue teacher, nor regulated educational pathways in Sweden, although several universities in Sweden do offer courses and programmes on different aspects of mother tongue instruction.In this presentation, an online resource for professional development for mother tongue teachers in Sweden will be presented, and analysed. This resource is the result of an initiative and funding from the Swedish National Agency for Education, and collaborations between researchers, mother tongue teachers and The National Agency for Education over one academic year. The online resource comprises seven units which each include a theoretical text exploring didactic or pedagogical perspectives relevant for mother tongue teachers, discussion questions addressing central concerns in the theoretical text and a classroom activity which mother tongue teachers are then invited to discuss, test and evaluate together with colleagues. The theoretical texts and activities were produced by a team that included both researchers and mother tongue teachers. In each of the seven units, perspectives from Sweden’s five national minority languages are also provided, given the very particular conditions under which students and teachers of these languages work.In the presentation, we argue that the ideological space created by policies that provide the opportunity to study mother tongues at school in Sweden stimulated the creation of this online resource, which opens up an implementational space for the development of mother tongue education (Hornberger 2005). The professional development programmes are moreover, created through collaboration across the scales of context, anchoring them in both20research and best practice, as well as providing a valuable space for interaction and learning for mother tongue teachers.

In this thesis, we investigate dependency parsing for commercial application, namely for future integration in a dialogue system. To do this, we conduct several experiments on dialogue data to assess parser performance on this domain, and to improve this performance over a baseline. This work makes the following contributions: first, the creation and manual annotation of a gold-standard data set for dialogue data; second, a thorough error analysis of the data set, comparing neural network parsing to traditional parsing methods on this domain; and finally, various domain adaptation experiments show how parsing on this data set can be improved over a baseline.  We further show that dialogue data is characterized by questions in particular, and suggest a method for improving overall parsing on these constructions. 

This study investigates the usefulness of the Treebank of Learner English (TLE) when applied to the task of Native Language Identification (NLI). The TLE is effectively a parallel corpus of Standard/Learner English, as there are two versions; one based on original learner essays, and the other an error-corrected version. We use the corpus to explore how useful a parser trained on ungrammatical relations is compared to a parser trained on grammatical relations, when used as features for a native language classification task. While parsing results are much better when trained on grammatical relations, native language classification is slightly better using a parser trained on the original treebank containing ungrammatical relations.

The article comprises an introduction to and an edition and translation of an Old Swedish sermon fragment found in the Hannaas Collection at the Ethno-Folkloristic Archive, University of Bergen, Norway (Hannaas 66). This previously unpublished paper fragment is one of the missing parts of the Old Swedish Homily Book (known as Svensk järteckens postilla), dating from the second half of the fifteenth century and now housed at the Royal Library in Stockholm, Sweden (Cod. Holm. A 111). The text in Hannaas 66 comprises a sermon for the 8th Sunday after Trinity based on Matthew 7. 15–16 and includes a miracle exemplum that illustrates the importance of acting justly and following the will of God.

The Danish National Archives in Copenhagen houses several thousand manuscript fragments, the remains of numerous works that were cut up and used in the bindings of later books. The majority of these fragments are written in Latin, Middle Low German, or Danish, although a few in Old Swedish also survive. Five of these Old Swedish fragments are published and discussed in this article. They contain parts of two of St Birgitta’s Revelations (Liber Caelestis) and of St Bernard’s A Rule of Good Life (Ad sororem modus bene vivendi in christianam religionem), known in Old Swedish as Ett gudhelikt lifwærne. The Birgittine texts are from an early stage of the retranslation process when compared to other extant versions and include several unique wordings that demonstrate the specific use of the original manuscript in a monastic environment. The Bernard fragments are one of just two extant versions and appear to predate the version in Stockholm, Royal Library, A 9; as such, they are an important witness to the propagation of the saint’s writings in Sweden.

The Danish National Archives in Copenhagen houses several thousand manuscript fragments, the remains of numerous works that were cut up and used in the bindings of later books. The majority of these fragments are written in Latin, Middle Low German, or Danish, although a few in Old Swedish also survive. Five of these Old Swedish fragments are published and discussed in this article. They contain parts of two of St Birgitta’s Revelations (Liber Caelestis and Revelationes Extravagantes) and of St Bernard’s A Rule of Good Life (Ad sororem modus bene vivendi in christianam religionem). The Birgittine texts are from an early stage of the retranslation process when compared to other extant versions and include several unique wordings that demonstrate the specific use of the original manuscript in a monastic environment. The Bernard fragments are one of just two extant versions and may predate the version in Stockholm, Royal Library, A 9; as such, they are an important witness to the propagation of the saint’s writings in Sweden.

De fleste studier om jøder, muslimer og kristne i middelalderens Europa fokuserer på de områder, hvor de levede sammen. Behandlingen af de ikke-kristne i de komplekse spændinger mellem kirken og de sekulære myndigheder danner kernen i disse studier, som fx sigter på at undersøge de spontane udbrud af antijødisk vold i dele af Central- og Vesteuropa eller virkeligheden bag den fredelige sameksistens, ”convivencia”, i Spanien. Følgerne af bl.a. korstogene, pesten, økonomisk nedtur samt politisk og social uro i disse områder regnes som de vigtigste faktorer i de skiftende forhold mellem de tre grupper naboer. Forholdene påvirkede også de skriftlige fremstillinger, da litteratur i middelalderen ligesom nu var et af de vigtigste midler, hvorved man udtrykte sin verdensopfattelse og skabte mening og struktur i det omkringliggende samfund. Disse tekster giver en et vigtigt indblik i, hvordan kristne europæere tænkte og hvad der optog dem.

The article examines the portrayal of Jews in medieval texts written in Danish before 1515. It begins by describing the theological basis for and creation of a ‘fantasy Jew’. The perception of Jews was fundamentally shaped by the idea that they had tortured and killed the Christian messiah. Devotional texts, sermons and Passion stories which describe the Jews as Christ killers are therefore discussed in detail, and the image of the deicide Jew in vernacular texts is shown to be malleable and changing. The image of the violent Jew who tortured and killed Jesus was used to arouse empathy among readers and to chastise them for being too like the Jews by behaving sinfully. Other Jewish ‘types’ that occur in the material are also investigated: effeminate, Satanic and usurious Jews as well as comparisons with animals. The preliminary results of an investigation into the type of language that was used to shape the image of Jews show that certain ‘negative’ words were used disproportionately more frequently in descriptions of Jews than of non-Jews. This suggests a powerful association between such words and the perception of Jews — a connection that was supported and further enhanced through religious art and theatre. The article concludes by noting what is missing in the extant material and what this might tell us about medieval Danish attitudes towards Jews.

Between the Viking Age and the Middle Ages, there was a noticeable change in relations between Scandinavia and the Islamic world – the sources point to a shift from travel and trade to hostility and war. Muslims did not settle in the North until the eighteenth century, and during the Middle Ages there was little contact between Scandinavians and ‘real’ Muslims. So how did Danes and Swedes imagine and describe this Other? Is there anything unusual or unexpected about the portrayal of Muslims? How does this image compare to that of the other great religious opponent, the Jew? By investigating East Norse devotional texts, travel literature, saints’ lives, romances and accounts of Ottoman warfare, this paper aims to draw out some of the major themes in medieval Scandinavian descriptions of Muslims and Islam.