Doctoral Theses / Doktorsavhandlingar Institutionen för svenska, flerspråkighet och språkteknologi

Permanent URI for this collectionhttps://gupea-staging.ub.gu.se/handle/2077/19517

Browse

Now showing 1 - 20 of 61

Argumentation and agreement: Annotating and evaluating Swedish corpora for argumentation mining
(2025-09-12) Lindahl, Anna
Argumentation occurs in all parts of life, and as such, is studied across disciplines. In natural language processing, the field of argumentation mining aims to develop computational tools that automatically analyze and evaluate argumentation. Such tools have many uses, from automatically grading essays to identifying fallacies. In order to build such tools, annotated data is essential both for training and evaluating, especially with large language models (LLMs). Creating annotated datasets, however, presents significant challenges, not only because of the complexity of argumentation but also methodological questions such as how to represent argumentation and how to evaluate annotation quality. To create more resources as well as investigate these challenges, in this thesis, I explore several approaches to argumentation annotation. To this end, I also present a comprehensive survey of argumentation annotation. Three annotation approaches of varying complexity are explored: argumentation schemes applied to editorials, argumentative spans to online forums and political debates, and attitude annotation to tweets. The datasets thus represent a wide variety of genres and approaches. Attitude in tweets was found to show the highest agreement among annotators, while annotation of editorials with argumentation schemes was the most challenging. In the evaluation of the annotations, several types of disagreement were identified. Most saliently, disagreement often occurred in cases where multiple interpretations are possible, challenging agreement as the primary measure of quality. These findings demonstrate the need for more comprehensive evaluation approaches. I therefore demonstrate ways to evaluate beyond single agreement measures: agreement analysis from multiple angles, annotator pattern investigation, and manual inspection of disagreement. To further explore argumentation annotation, I investigate how two different LLMs annotate argumentation compared to human annotators, finding that while the models exhibit similar annotation behavior as humans, with similar agreement levels and disagreement patterns, the models agree more among themselves than human annotators.
The South in the North: Raciolinguistic Perspectives on Family Language Policy in Sweden
(2025-08-25) Yeshalem, Abraham
The thesis examines the FLP-making practices and multilingual familial experiences of some Southern migrant families in Sweden. Specifically, the thesis investigates (1) children’s voices and perspectives on their multilingual repertoire and multilingual familial experiences; (2) how family members experience the interplay between the majority and minority languages in their daily familial interactions, and (3) how individual FLP decisions and practices interact with the broader language-related discourses and ideologies in the majority society. The thesis employs multipronged methodological approaches, including language portrait methods of body mapping, space-mapping, post-mapping narration, semi-structured interviews, participant observation, and media text analysis. The thesis is based on three articles, and the analysis draws on a multi-layered framework, which brings notions and approaches from different analytical traditions, such as child agency, language ideology, raciolinguistic ideologies and raciolinguistic perspectives, and critical discourse analysis, into productive dialogue. The results indicate that FLP experiences are filled with ambivalence and language choice dilemmas, influenced by both family-internal factors (such as family constellation, language proficiency asymmetry, competing linguistic demands and interests, and child agency), and also by family-external factors, namely racialized perceptions of languages and standard language ideology that appear to be internalized and enacted by family members, which affect their FLP decisions and language use practices. Overall, the results reveal that the process of language socialization in minority languages is not merely a matter of a private FLP decision that can be practiced based on individual dispositions and beliefs; rather, it appears to be shaped, in part, by prevailing language ideologies and identity-related discourses in society. This could be because communicative practice in a minority language is not only a means of fostering identity expression and continuity in the heritage language and culture, but also a linguistic practice contributing to the social positioning of otherness. Ultimately, the theoretical argument of the thesis is that understanding the FLP experience of Southern families in the Global North needs critical and social justice lenses since they are situated in a context where their racial and linguistic identities come into play against the monoracial culture of a monoglossic standard.
Den unga skånskan. Dialekt och platsanknytning hos gymnasieungdomar i nordvästra Skåne.
(2025-05-12) Horn, Greta
This dissertation aims to contribute to research on young people’s dialect use in contemporary urban environments. In this study, quantitative and qualitative linguistic methods are used to in- vestigate, (both in terms of form and content), the dialect use of a group of twenty-one adolescents in northwest Scania. Their realization of dialect and standard variants of twelve selected linguistic variables is mapped. A measurement of dialectal use obtained from this survey is then related to the informants’ degree of attachment to the place where they live. The result of the linguistic variation is operationalized through a dialect index. The linguistic variation is then related to place attach- ment by analyzing interview questions and using an index of local attachment. I have also investi- gated if the linguistic variation is affected by other non-linguistic variables, such as gender, where one lives, how one talks about one’s place of residence, and the attitude towards one’s dialect. The results show that one group of informants is both dialect-speaking and has a high local attachment score. One conclusion regarding the linguistic variation is that the standard feature retroflexion seems quite frequent in the material. Still, only the most standard language-speaking informants use this particular feature. Retroflexion and standard language alveolar r are prominent signs of dialect leveling in this study. In the discussions regarding staying or moving in the future, i.e. mental mobility, it is my im- pression that several informants are trying to conform to some kind of mobility norm. Those who are convinced that they want to stay do additional interactional work by adding laugh or discourse markers to show an awareness that their preferred position is not necessarily the expected stance. I discuss this in relation to the new mobilities paradigm. The young Scanian is the title of this dissertation. The term could be used to describe the Sca- nian spoken by young people in Scania today, but in this dissertation, it is also a quote from an informant who says this is the variety of her generation: the young Scanian. Several informants sta- te that they and their peers speak this new, young variety of Scanian. For the young people in this study, dialect seems to be something belonging to the past, not to the young generation. For them, the young Scanian is something other than “dialect” and can be understood as a variety of its own.
Litteratur på omvägar. Tre perspektiv på indirekta romanöversättningar till svenska 2000–2015
(2025-03-18) Allwood, Anja
Indirect translation, translating from another translation, has been an important part of intercultural communication for millennia. Yet, the ways in which it functions are still not fully understood. This thesis sets out to explore indirect translation (ITr) in the context of present-day literary translation in Sweden. The thesis is structured around three empirical studies. The first study establishes the original source languages and mediating languages of ITrs into Swedish 2000–2015. The second study is based on a questionnaire with the publishers of ITrs, where their views on ITr are investigated. The third study is a textual comparison between an original Afrikaans-language novel, its intermediate translation into English, and the indirect translation into Swedish. The results show that while the number of indirect translations is modest, the variety of ultimate source languages and cultures contributes to the overall diversity of the Swedish literary market. The publishers are aware of the norm of direct translation and juggle carefully between different options, choosing ITr when relevant. The study also shows that it is possible for an indirect translation to follow the original quite closely, albeit with some toning down of specific features of the text. All in all, the thesis increases our understanding of indirect translation and the ways in which it functions in present-day Sweden, contributing thus to a wider global grasp of the phenomenon in question.
Locating, Constructing, and Disciplining Self and Other: A Discourse Ethnography of Civic Orientation in Sweden
(2025-02-17) Bauer, Simon
This thesis is a multilevel discourse ethnography concerned with discourses and discursive constructions pertaining to Civic Orientation for Newly Arrived Migrants in Sweden. Civic Orientation is a provision that follows a general trend often referred to as ‘the civic turn’, which describes how states in the Global North have converged in their migration governance by using educational programmes as the primary tool through which they aim to ‘integrate’ migrants. The programme, established in 2010, aims to facilitate entry into the labour market and society. Taught in the participants’ first language, or another language that they have good command of, the programme is a rich site of multilingual interaction and politics. Drawing on a range of theoretical concepts, but anchored in a Foucauldian research tradition, this thesis analyses (1) policy documents regulating the programme; (2) a print media corpus covering publications about civic orientation between 2002 and 2021; (3) individual semi-structured interviews with 14 people charged with interpreting and implementing the policy; and (4) classroom ethnographic fieldnotes from six such courses delivered during the first half of 2020, three each in English and Arabic. The analysis focuses in particular on (1) how values are discursively negotiated and constructed; (2) the construction of Sweden as a nation-state in relation to the migrants’ countries of origin, and (3) how specific ways of being Swedish and non-Swedish manifest throughout the programme. The thesis is based on four articles, and, through a multilevel analysis, the results show how values such as gender equality and democracy are nationalised as specifically Swedish values. Such constructions can be seen from the print media corpus, through the interpretation of policy documents, and in practice in the classrooms. Furthermore, Sweden is presented as the best country in the world in terms of values and how to organise society, in contrast to the Arab World in particular. Through the deployment of common-sense discourses based in Swedish exceptionalism, Sweden is presented as the pinnacle of scientific and human progress. Such a discourse further plays a crucial part in the construction of Swedishness, through which migrants’ knowledge(s) and experience(s) is used to make them realise how and why they are wrong and how they can change into ‘developed Swedes’. Ultimately, the thesis illustrates how state power is exercised discursively and materially through civic orientation.
In the minds of stochastic parrots: Benchmarking, evaluating and interpreting large language models
(2024-11-18) Morger, Felix
The arrival of large language models (LLMs) in recent years has changed the landscape of natural language processing (NLP). Their impressive performance on popular benchmarks, ability to solve a range of different tasks and their human-like linguistic interactional abilities, have prompted a debate into whether these are just "stochastic parrots" who are cleverly repeating what humans say without understanding its meaning or whether they are acquiring essential language capabilities, which would be an important stepping stone towards artificial general intelligence. To tackle this question, developing analysis methods to measure and understand the language capabilities of LLMs has become a defining challenge. These include developing benchmarks to reliably measure their performance as well and interpretability methods to gauge their inner-workings. This is especially relevant at a time when these models already are having a considerable impact on our society. An increasing amount users are affected by the technology and calls are made for transparent, regulated and thorough evaluation of AI. In these efforts, it is important to estimate the possibilities and limitations of these analysis methods since they will play an important role in holding technologies in AI accountable. In this compilation thesis, I expound on the components and processes involved in analyzing LLMs. The articles included in this compilation thesis use different approaches for analyzing LLMs, from introducing a multi-task benchmark Superlim for Swedish NLU to investigating LLMs' ability to predict language variation. To this effort I explore what the possibilities and limitations are of popular analysis methods and what implications these have for developing LLMs. I argue that integrating explanatory approaches from empirical linguistic research is important to understand the role of both the data and the linguistic features used when analyzing LLMs. Doing so does not only help guide the development of LLMs, but also bring insights into linguistics.
Språkkunskaper och kunskapsspråk. Ämneslitteracitet i ämnesprov, undervisning och elevtexter inom svenska som andraspråk och fysik
(2024-10-02) Håkansson, Camilla
The overall aim of this thesis is to gain a deeper understanding of disciplinary literacy in two subjects, Swedish as a second language (SSL) and physics, among ninth-year students in compulsory school. This thesis focuses on students learning subjects in a second language and analyses disciplinary literacy from three perspectives. It investigates how literacy expectations are expressed in two national tests for SSL and physics, how disciplinary literacy is explicitly utilised in collaboratively teaching of these subjects and how students articulate their subject knowledge in their final assignment. The data include national tests, classroom observations in SSL and in physics, teaching materials and students’ texts. This thesis is grounded in empirical studies, primarily using Legitimation Code Theory and Systemic Functional Linguistics to analyse and interpret the data. The results indicate that national tests in both SSL and physics have high literacy expectations, but they manifest differently. In Swedish, literacy expectations are characterised by high demands for reading and writing, requiring students to read various genres and write longer texts. The vocabulary is mainly not subject specific. In contrast, the national test in physics demands disciplinary vocabulary and subject-specific texts but expects fewer and shorter texts. The results of the classroom study reveal that disciplinary literacy is addressed in both subjects, but to a limited extent. The teaching of SSL primarily focuses on general school literacy, with some emphasis on metalanguage that indicates disciplinary literacy. In physics, the teacher emphasises subject-related concepts and explains scientific reasoning, but only to a small extent. This thesis also analyses students’ texts. The results reveal that most of the students adhere to the macrostructure of an argumentative text, as they are explicitly taught. However, the paragraph level is challenging for the students. This indicates a need for more in-depth knowledge of text structure and the linguistic resources required for more advanced texts. In particular students who learn subjects in a second language need to be taught not only general school literacy but also subject-specific language in connection with subject knowledge.
Tala fram texten: När barn med läs- och skrivsvårigheter skriver med tal-till-text
(2023-05-22) Kraft, Sanna
Effective writing is a crucial skill that requires not only the mastery of various sub-processes but also deliberate orchestration of those sub-processes within the constraints of limited workingmemory capacity. In practice, some of those sub-processes, such as transcription (spelling and handwriting), need to be automatised to free up capacity for other processes. Unfortunately, children who struggle with spelling rarely manage to fully automatise transcription. This often has a negative impact both on their writing process and on the final text. One solution to this problem could be to let them use speech-to-text (STT), because this would allow them to avoid spelling by using their voice to create text. This thesis consists of four articles where I investigate whether, and if so how, composing by means of STT can facilitate writing for children (aged 10–13) with reading and writing difficulties. I analyse both the composition processes and the final texts. I make comparisons both with text production by means of a keyboard and with a reference group of children without difficulties. In addition, I explore whether and how successful use of STT correlates with individual linguistic and cognitive skills. Overall, my results suggest that STT can indeed facilitate some aspects of writing for children with reading and writing difficulties, though not emphatically. In my studies, the use of STT does not yield any improvements at group level in processes such as meaning-related revisions or in assessed text quality. However, one important caveat is that the participants received only a very short introduction to STT. In general, the ability to use a new tool effectively is of course likely to improve with instruction and practice. Importantly, my results do suggest that instruction in STT use in conjunction with writing processes such as revising is crucial for successful usage of this tool.
Neutrala substantiv på -ande i text och ordbok
(2022-08-25) Holmer, Louise
This thesis presents an examination of Swedish deverbal neutral nouns with the suffix ande in a corpus of texts from the 1900s and early 2000s, compared to the presentation of these deverbal nouns in three modern, monolingual Swedish dictionaries. The aim of this dual study is to situate the usage of -ande nouns within a lexicographical perspective. In the corpus study, I examine deverbal nouns ending with ande and their frequency in several different genres, including news texts, academic texts, novels, and online blogs. The most frequent ande nouns (included in five different sub-corpora) are further examined with regard to their semantics and contextual properties. The corpus study shows that academic texts display the highest relative frequency of nouns ending in ande and blog texts the lowest frequency. The specialized corpus studies show that highly frequent ande-derivatives often display polysemy, and that they may be combined with several different adjectives, depending on each sub-sense, respectively. Regarding the dictionaries used in this study, I investigate deverbal neutral nouns with the suffix ande and their lexicographical realization on three structural levels: (i) in the prefatory texts of each dictionary, (ii) in the set of headwords, and (iii) in the entry texts. The dictionaries investigated are Bonniers svenska ordbok (BSO 2010), Svenska Akademiens ordlista över svenska språket (SAOL 14, 2015), and Svensk ordbok utgiven av Svenska Akademien (SO 2009). The results of this investigation show that SAOL 14 contains considerably more words ending with ande in its list of headwords than the other two dictionaries. BSO 2010 treats many of its headwords that end with ande as a kind of mixed category. The words are listed, but a semantic description is not always provided. In contrast, SO 2009 lists a majority of the ande derivatives in the article text that accompanies the root verb. The results from the combined investigations show that the most highly frequent words display varied contexts in the corpora, something that is not always picked up by the dictionaries. Another result is that polysemous ande words, and ande words that form the basis of certain phrases, are to a great extent represented in the dictionaries. In light of the corpus investigations, I argue that all three dictionaries could be more consistent in their treatment of derived words. The semantic descriptions and the formal and contextual descriptions presented in these dictionaries may also benefit from a more ‘usage-based’ revision. I also identify considerable differences between the print dictionaries and the electronic versions produced by individual publishers.
Inte för räddhågsna. Undervisning i grundläggande litteracitet och svenska som andraspråk på gymnasieskolans språkintroduktion
(2021-08-16) Winlund, Anna
The overall aim of this thesis is to gain a deeper understanding about the education of Swedish as a second language and basic literacy to recently immigrated adolescents with little prior experience of school-based learning. It investigates how the students are given access to the literacy practices that are required for active and independent participation in school and in society, how the students take part in the activities and interactions offered in this particular school context, how the relationship between the teachers and these adolescents is manifested and finally, how the rules of the school context relate to the literacy practices that are available to the students. The thesis is based on four ethnographic studies investigating diverse aspects of this education, using different theoretical frameworks. The data consists of observations of lessons in Swedish and social sciences, conversations with the teacher, the students and the language tutor, in addition to formal interviews with the students. The results indicate that the teacher and the language tutor played an important role to give the students access to the literacy practices and rules of schooling. Students’ previous knowledge, as well as class field trips and concrete examples, served as important foundations for their instruction. Also, the teachers’ engagement with the students’ linguistic and other semiotic resources contributed to the students’ participation in literacy practices. Concurrently, the interaction about topics that had previously been unfamiliar to the students seemed to enhance their ability to understand new discourses. The study implies that some of the rules and norms associated with schooling promote learning in this specific context, while other rules seemed adapted to prepare the students for future studies. The thesis also discusses students’ agency and opportunities to invest in their schooling in the mainstream society.
Learning of Definiteness by Belarusian Students of Swedish as a Foreign Language
(2021-04-28) Agebjörn, Anders
Through a series of studies, this thesis investigates the learning of definiteness in Russian-speaking students of Swedish. A communicative, oral-production task elicited modified and non-modified noun phrases in indefinite and definite contexts. Study I describes the development of the morphosyntactic structure through which Swedish encodes definiteness, the association between this structure and its meaning, and the relationship between those two tasks over time. Using an English version of the elicitation task and a test of metalinguistic knowledge, Study II examines the relationship between the learners’ explicit knowledge of article semantics and their actual use of English articles. Adding a test of language-learning aptitude, Study III then explores both the influence of second-language English and that of aptitude on the development of Swedish. Finally, Study IV discusses the role of complexity and input frequency. The main findings include that, at the onset of Swedish study, the learners had minimal knowledge of the morphosyntactic structure but were generally sensitive to the meaning of definiteness. However, knowledge of form developed over time while knowledge of meaning did not, and the two learning tasks did not appear to be directly related to each other. In addition, the learners were seldom aware that choosing between indefinite and definite articles require the speaker to take the hearer’s perspective, but this lack of metalinguistic understanding did not seem to affect their use of articles. Further, previous knowledge of English appeared to facilitate the development of a Swedish morpheme that is structurally similar to its English counterpart, while aptitude was associated with the development of a morpheme whose English counterpart is structurally different. Finally, the learners used high-frequency morphemes more consistently than low-frequency ones, and morphemes were more likely to be supplied in frequent constructions than in infrequent ones. These findings are discussed in relation to a modular, cognitive framework for language learning and use.
Exploring natural language processing for single-word and multi-word lexical complexity from a second language learner perspective
(2021-02-09) Alfter, David
In this thesis, we investigate how natural language processing (NLP) tools and techniques can be applied to vocabulary aimed at second language learners of Swedish in order to classify vocabulary items into different proficiency levels suitable for learners of different levels. In the first part, we use feature-engineering to represent words as vectors and feed these vectors into machine learning algorithms in order to (1) learn CEFR labels from the input data and (2) predict the CEFR level of unseen words. Our experiments corroborate the finding that feature-based classification models using 'traditional' machine learning still outperform deep learning architectures in the task of deciding how complex a word is. In the second part, we use crowdsourcing as a technique to generate ranked lists of multi-word expressions using both experts and non-experts (i.e. language learners). Our experiment shows that non-expert and expert rankings are highly correlated, suggesting that non-expert intuition can be seen as on-par with expert knowledge, at least in the chosen experimental configuration. The main practical output of this research comes in two forms: prototypes and resources. We have implemented various prototype applications for (1) the automatic prediction of words based on the feature-engineering machine learning method, (2) language learning applications using graded word lists, and (3) an annotation tool for the manual annotation of expressions across a variety of linguistic factors.
Det-konstruktioner i bruk. En systemisk-funktionell analys av satser med icke-referentiellt det i modern svenska
(2020-09-24) Thyberg, Kajsa
The aim of this thesis is to map out functions of modern Swedish det-constructions, i.e. different clausal patterns including the non-referential third person pronoun det, for example, det är kul att spela tennis (‘it’s fun to play tennis’), det fläktade en skön vind från sjön (‘there was a nice wind blowing from the lake’) and det är mysigt med blommor och växter i hemmet (‘it’s cosy with flowers and plants at home’). The research questions concern how the world is semantically represented, how assessment can be expressed and how non-referential det is distributed in comparison to referential det. The study uses corpus data consisting of sentences with det in texts from news reports, personal blogs and academic works. The analytical perspective derives from Systemic Functional Grammar, focusing on the ideational and interpersonal metafunction. Basic concepts used from the theory are process, modality and modal assessment. The results show that non-referential det has approximately the same frequency as referential det in all three investigated text types, and that referential det sometimes has a wide and less clear kind of reference. Most of the det-constructions operate as a relational process where the first participant may be either implicit or construed as a medium as the process is non-agentive. However, passive det-constructions differ considerably from active ones as the process implicates an agent to a much higher degree – yet without expressing the agent role. Furthermore, the process of the passive det-construction is not relational but mental, verbal or material. The results also show that det-constructions can be used as a modal resource expressing several types of modality and modal assessment. This interpersonal grammar is regularly construed metaphorically by a relational clause as a semantically extending alternative to an adjunct or finite construction. However, some det-constructions are used as a modal resource which is considered not metaphorical, as they do not have the adjunct/finite alternative. In addition, the borderline between metaphor and non-metaphor regarding the constructions is investigated and discussed. By adopting a systemic-functional perspective on Swedish det-constructions, this study not only contributes a deeper understanding of the functions, but also increases the description of the Swedish language in systemic-functional terms, thereby developing the theoretical application.
Från beslut till broschyr. Intertextualitet, äldre och kultur i texter inom en statlig satsning
(2020-01-20) Sandberg, Malin
The overarching purpose of this study is to map out how intertextual relations are manifested in texts produced within a government initiative on arts for the elderly, and to shed light on how discourses on the elderly and the arts are articulated in texts from the initiative. This purpose is addressed with qualitative research methods, and with a focus on three themes: intertextual relations, representations of the elderly, and constructions of the purposefulness of the initiative. The study is regarded as a case study, which at a more general level can contribute to an understanding of texts in institutional contexts as well as public discourses on the arts and the elderly. The study draws on data consisting of a collection of texts produced within the three-year initiative called 'Kultur för äldre' ('Arts for the elderly'), which was launched in 2011 and aimed to increase elderly peoples’ access to the arts. The texts are written by different authorities and professionals: the Swedish government, the Swedish Arts Council, local project organizers, and an external consultant. They vary by genre, ranging from government decisions to information brochures, and fill various purposes in different, but related, contexts. The results show that there are both explicit and implicit intertextual relations between the texts in the data, and that three main intertextual chains can be defined. These are traceable from explicit markers of intertextuality, as well as contextual information and by following certain phrases that are recontextualized between texts. Analyses of representations of social actors, constructions of purposefulness, and visual resources in the data, show that discourses on the elderly as dependent and in need of health care tend to be both reflected and reproduced in the data. The arts, finally, tend to be framed within a social welfare discourse, and reproduced as important, first and foremost through utilization rather than in its own right.
Etableringen av ha-bortfall i svenskan. Från kontaktfenomen till inhemsk konstruktion.
(2019-12-19) Bäckström, Linnéa
This thesis presents an investigation into the causes surrounding the emergence and spread of temporal auxiliary verb omission in Swedish. In modern Swedish, certain finite auxiliary verbs can be elided in the perfect and the pluperfect in subordinate clauses. Furthermore, the auxiliary verb in non-finite perfects can be omitted (subject to a number of restrictions) in three different classes of non-finite clause. At the end of the 17th century, auxiliary verb omission spread rapidly in all types of subordinate clause. It has been suggested in previous research that this might be the result of the influence of German. The emergence and spread of non-finite auxiliary verb omission has not been previously examined. Traditionally, it has been suggested that omission of the non-finite auxiliary verb was a grammatical feature of the language since the Old Swedish period. In this thesis, I report on two different studies. In the first, I investigated the course of events outlining the spread of finite auxiliary verb omission during the Early Modern Swedish period. In the second study, I examined the occurrence of infinitive perfects in Old- and Early Modern Swedish. Additionally, I searched for instances of auxiliary verb omission in those very constructions. The results of the study of the finite auxiliary verbs were in agreement with previous studies in this area of grammar, as expected. The results of the study of the non-finite verb forms, however, were not anticipated: non-finite auxiliary verb omission spread throughout the language only after the finite auxiliary verb omission was already established. Taking these results into consideration, I assume that finite auxiliary verb omission emerged in Swedish due to the process of grammatical replication from New High German. The (optional) possibility of auxiliary verb omission then spread to infinitival clauses by analogical extension. A number of consequences of auxiliary omission are the emergence of the supine, a specific perfect participle form which differs in form and meaning from the past participle and the loss of temporal auxiliary verb form vara, ‘be’.
Splitting rocks: Learning word sense representations from corpora and lexica
(2019-08-23) Nieto Piña, Luis
The representation of written language semantics is a central problem of language technology and a crucial component of many natural language processing applications, from part-of-speech tagging to text summarization. These representations of linguistic units, such as words or sentences, allow computer applications that work with language to process and manipulate the meaning of text. In particular, a family of models has been successfully developed based on automatically learning semantics from large collections of text and embedding them into a vector space, where semantic or lexical similarity is a function of geometric distance. Co-occurrence information of words in context is the main source of data used to learn these representations. Such models have typically been applied to learning representations for word forms, which have been widely applied, and proven to be highly successful, as characterizations of semantics at the word level. However, a word-level approach to meaning representation implies that the different meanings, or senses, of any polysemic word share one single representation. This might be problematic when individual word senses are of interest and explicit access to their specific representations is required. For instance, in cases such as an application that needs to deal with word senses rather than word forms, or when a digital lexicon's sense inventory has to be mapped to a set of learned semantic representations. In this thesis, we present a number of models that try to tackle this problem by automatically learning representations for word senses instead of for words. In particular, we try to achieve this by using two separate sources of information: corpora and lexica for the Swedish language. Throughout the five publications compiled in this thesis, we demonstrate that it is possible to generate word sense representations from these sources of data individually and in conjunction, and we observe that combining them yields superior results in terms of accuracy and sense inventory coverage. Furthermore, in our evaluation of the different representational models proposed here, we showcase the applicability of word sense representations both to downstream natural language processing applications and to the development of existing linguistic resources.
Flerspråkiga elevers läsförståelse på svenska. Om lässtrategier och läsutveckling på mellanstadiet
(2019-05-21) Lindholm, Anna
The overall aim of this thesis is to highlight the significance of reading strategies in developing multilingual middle-school students’ Swedish reading comprehension. Reading strategies are considered from teaching, reading development and individual perspectives, and this compilation thesis comprises an introductory chapter and three articles. The three empirical studies are based on material collected at a school where around 95% of students are multilingual. Overall, a Mixed Methods Research (MMR) approach was taken. The first substudy focuses on the teaching of reading strategies in three different subjects through observations of two multilingual classrooms as well as teacher interviews. The second substudy examines students’ reading development over two years, and its relation to their self-reported use of reading strategies. The third substudy investigates the reading comprehension of eight multilingual students and their reflections on the use of reading strategies. The studied teachers have been inspired by the Reciprocal Teaching model, but the results indicate that these strategies (summarizing, predicting, clarifying, and questioning) are not always the ones students need in their own reading. In individual reading situations, students successfully tackled the vocabulary problems that arose when they read subject content by rereading sentences or paragraphs or drawing on their existing knowledge. Nevertheless, some students experienced difficulties comprehending the overall text content.
”See there I’m stuck now!” Samtalsdeltagares orientering mot att tala svenska i amerikasvenska dialektintervjuer
(2019-04-10) Adamsson Eryd, Henrietta
This thesis is a qualitative investigation of a communicative project called tala svenska (speak swedish) in American Swedish dialect interviews. The aim of the study is to analyze and describe the communicative project. Based on this aim the study investigates the interviews as activities with accompanying frames, roles and phases. Interactional resources and how those are being used in doing the communicative project, and the activity, are also a part of the analysis. The material consists of audio and video recordings of American Swedish dialect interviews and the analysis is based on 12 recordings with 10 interviewees/participants and 6 interviewers. The participants are of different ages and have different reasons for speaking Swedish. Some of them have visited Sweden and speak Swedish on a regular basis, while some do not have anyone to talk Swedish to and have never visited Sweden. The analysis of the interviews, and the speakers’ usage of the interactional resources, show that in doing the joint communicative project of speaking Swedish, the participants and the interviewers use different strategies. Based on activity roles, the participants orient themselves toward achieving speaking Swedish, while the interviewers act toward the participants’ speaking by using intentional passivity. In the American Swedish dialect interviews the communicative project, i.e. speaking Swedish, is made visible through the participants use of the interactional resources code alternation, repair, meta comments and metalinguistic comments. Through marked and unmarked code alternation and through side sequences made by code alternation the participants orient themselves toward the communicative project of speaking Swedish as a part of the activity American Swedish dialect interviews. Repair and metalinguistic comments are being used to mark code alternation and that makes the communicative project tala svenska visible by speakers focusing on speaking Swedish. Code alternated side sequences that are meta comments or side sequences that do a different activity than the ongoing dialect interview, make the communicative project tala svenska visible by speakers uttering them in another code, i. e. leaving the inner frame and inner activity both by action and code.
Strategic Vocabulary Learning in the Swedish Second Language Context
(2019-02-22) LaBontee, Richard
This large-scale research project represents an exploratory investigation into the reported vocabulary learning strategies (VLS) used by adult, beginner Swedish L2 learners living and studying in Sweden. A questionnaire instrument, the Swedish Vocabulary Learning Strategy Survey (SVLSS) built explicitly for data collection in this context is developed and used over the course of five studies regarding learners’ approaches to Swedish L2 vocabulary learning. Results from the first study are used to establish a preliminary item list for the SVLSS from collected interview and learning task data. Pilot results (SVLSS 1.0, 1.1) guide revisions to access- ibility, readability, and item list, resulting in a 74-item questionnaire (SVLSS 1.2). The second study adopts a six-category VLS taxonomy for the instrument that is extracted through the guidance of exploratory factor analysis. Findings are used to conduct revisions aimed at supporting the adopted taxonomy, and again to improve accessibility, and readability. The third study situates the SVLSS instrument within a comparative review of other VLS questionnaires, guiding extended revisions started in study two. Revision results in the acceptance of an updated VLS taxonomy, and in the 69-item SVLSS (2.0). The fourth study explores what the target demographic believes it means ‘to know a word’ as a means of better these learners’ vocabulary learning experience. The fifth study uses the SVLSS 2.0 to explore possible patterns in learners’ VLS use across demographic grouping variables, offering two emergent learner profiles. Findings across these studies indicate that adult, beginner learners of Swedish L2 vocabulary report using strategies for establishing new word information more than any other VLS type, suggesting that the need to acquire vocabulary knowledge before it can be used in other strategic manners is high for this demographic. Also, significant differences between learners’ use of VLS are seen even amongst relatively minor differences in learners’ beginner proficiency levels, adult age groups, and amounts of time spent learning the language. A synthesis of findings suggest that these learners value communicative practices for learning words, though may not be able to reflect this in their learning behavior at earlier levels of Swedish. This report concludes with suggested use guidelines and planned updates for the SVLSS instrument, as well as suggested and planned future research for the field of Swedish L2 VLS use.
Skriftlig lärarrespons för vuxna nybörjare i svenska som andraspråk: teoretiska perspektiv, responspraktik och uppfattningar
(2019-02-14) Jakobson, Liivi
This compilation thesis serves to fill a research gap in teacher written feedback on second language (L2) writing and pronunciation by focusing on adult beginners of culturally diverse backgrounds. Basically, this thesis seeks to answer three questions: 1. What do teachers focus on and in what manner is feedback given?, 2. What are the teachers’ beliefs about feedback?, and, 3. What are the students’ preferences concerning feedback? The study presented in the first article concerns international students and elaborates on a model for analyzing teacher feedback practices regarding writing and pronunciation and the students’ ranking of that feedback. Based on the need for a holistic perspective recognized in the first study, the second article presents and discusses the research gap regarding feedback in adult beginner-learner L2 contexts and proposes a hermeneutic approach in the field of L2 research. In article three and four, a new feedback analysis model based on hermeneutics was applied for the design of a questionnaire, which includes the Likert scale and ranking. These studies investigated teachers’ (article three) and students’ (article four) perceptions and priorities regarding feedback. Data from the students’ questionnaires indicated that the students evaluated feedback on grammatical structures in writing as top priority. They generally valued all feedback manners, but clearly preferred specific praise and specific criticism. Results from the teachers’ questionnaires, combined with interviews, showed that they evaluated almost all categories as important but the ranking showed that there was a variation between teachers’ perceptions, which was influenced by a lack of time, and communication with the students, together with technical problems in web-based contexts. The analysis indicates several factors contributing to the complexity of feedback and, thus, providing feedback requires intuition and dexterity on the part of teachers. It is suggested that aligning students’ expectations and teachers’ practices and beliefs is especially important in distance-learning contexts in order to facilitate new understanding for clarifying the meaning of feedback.

Browse

Recent Submissions