| 1. |
- af Geijerstam, Åsa, docent, 1972-
(author)
-
Att skriva i naturorienterande ämnen i skolan
- 2006
-
Doctoral thesis (other academic/artistic)abstract
- When children encounter new subjects in school, they are also faced with new ways of using language. Learning science thus means learning the language of science, and writing is one of the ways this is accomplished. The present study investigates writing in natural sciences in grades 5 and 8 in Swedish schools. Major theoretical influences for these investigations are found within the socio-cultural, dialogical and social semiotic perspectives on language use.The study is based on texts written by 97 students, interviews around these texts and observations from 16 different classroom practices. Writing is seen as a situated practice; therefore analysis is carried out of the activities surrounding the texts. The student texts are analysed in terms of genre and in relation to their abstraction, density and use of expansions. This analysis shows among other things that the texts show increasing abstraction and density with increasing age, whereas the text structure and the use of expansions do not increase.It is also argued that a central point in school writing must be the students’ way of talking about their texts. Analysis of interviews with the students is thus carried out in terms of text movability. The results from this analysis indicate that students find it difficult to talk about their texts. They find it hard to express the main content of the text, as well as to discuss it’s function and potential readers.Previous studies argue that writing constitutes a potential for learning. In the material studied in this thesis, this potential learning tool is not used to any large extent. To be able to participate in natural sciences in higher levels, students need to take part in practices where the specialized language of natural science is used in writing as well as in speech.
|
|
| 2. |
- Basirat, Ali, 1982-
(author)
-
Principal Word Vectors
- 2018
-
Doctoral thesis (other academic/artistic)abstract
- Word embedding is a technique for associating the words of a language with real-valued vectors, enabling us to use algebraic methods to reason about their semantic and grammatical properties. This thesis introduces a word embedding method called principal word embedding, which makes use of principal component analysis (PCA) to train a set of word embeddings for words of a language. The principal word embedding method involves performing a PCA on a data matrix whose elements are the frequency of seeing words in different contexts. We address two challenges that arise in the application of PCA to create word embeddings. The first challenge is related to the size of the data matrix on which PCA is performed and affects the efficiency of the word embedding method. The data matrix is usually a large matrix that requires a very large amount of memory and CPU time to be processed. The second challenge is related to the distribution of word frequencies in the data matrix and affects the quality of the word embeddings. We provide an extensive study of the distribution of the elements of the data matrix and show that it is unsuitable for PCA in its unmodified form.We overcome the two challenges in principal word embedding by using a generalized PCA method. The problem with the size of the data matrix is mitigated by a randomized singular value decomposition (SVD) procedure, which improves the performance of PCA on the data matrix. The data distribution is reshaped by an adaptive transformation function, which makes it more suitable for PCA. These techniques, together with a weighting mechanism that generalizes many different weighting and transformation approaches used in literature, enable the principal word embedding to train high quality word embeddings in an efficient way.We also provide a study on how principal word embedding is connected to other word embedding methods. We compare it to a number of word embedding methods and study how the two challenges in principal word embedding are addressed in those methods. We show that the other word embedding methods are closely related to principal word embedding and, in many instances, they can be seen as special cases of it.The principal word embeddings are evaluated in both intrinsic and extrinsic ways. The intrinsic evaluations are directed towards the study of the distribution of word vectors. The extrinsic evaluations measure the contribution of principal word embeddings to some standard NLP tasks. The experimental results confirm that the newly proposed features of principal word embedding (i.e., the randomized SVD algorithm, the adaptive transformation function, and the weighting mechanism) are beneficial to the method and lead to significant improvements in the results. A comparison between principal word embedding and other popular word embedding methods shows that, in many instances, the proposed method is able to generate word embeddings that are better than or as good as other word embeddings while being faster than several popular word embedding methods.
|
|
| 3. |
- Björk, Ingrid, 1961-
(author)
-
Relativizing linguistic relativity : Investigating underlying assumptions about language in the neo-Whorfian literature
- 2008
-
Doctoral thesis (other academic/artistic)abstract
- This work concerns the linguistic relativity hypothesis, also known as the Sapir-Whorf hypothesis, which, in its most general form claims that ‘lan-guage’ influences ‘thought’. Past studies into linguistic relativity have treated various aspects of both thought and language, but a growing body of literature has recently emerged, in this thesis referred to as neo-Whorfian, that empirically investigates thought and language from a cross-linguistic perspective and claims that the grammar or lexicon of a particular language influences the speakers’ non-linguistic thought.The present thesis examines the assumptions about language that underlie this claim and criticizes the neo-Whorfian arguments from the point of view that they are based on misleading notions of language. The critique focuses on the operationalization of thought, language, and culture as separate vari-ables in the neo-Whorfian empirical investigations. The neo-Whorfian stud-ies explore language primarily as ‘particular languages’ and investigate its role as a variable standing in a causal relation to the ‘thought’ variable. Tho-ught is separately examined in non-linguistic tests and found to ‘correlate’ with language.As a contrast to the neo-Whorfian view of language, a few examples of other approaches to language, referred to in the thesis as sociocultural appro-aches, are reviewed. This perspective on language places emphasis on prac-tice and communication rather than on particular languages, which are vie-wed as secondary representations. It is argued that from a sociocultural per-spective, language as an integrated practice cannot be separated from tho-ught and culture. The empirical findings in the neo-Whorfian studies need not be rejected, but they should be interpreted differently. The findings of linguistic and cognitive diversity reflect different communicational practices in which language cannot be separated from non-language.
|
|
| 4. |
- de Lhoneux, Miryam, 1990-
(author)
-
Linguistically Informed Neural Dependency Parsing for Typologically Diverse Languages
- 2019
-
Doctoral thesis (other academic/artistic)abstract
- This thesis presents several studies in neural dependency parsing for typologically diverse languages, using treebanks from Universal Dependencies (UD). The focus is on informing models with linguistic knowledge. We first extend a parser to work well on typologically diverse languages, including morphologically complex languages and languages whose treebanks have a high ratio of non-projective sentences, a notorious difficulty in dependency parsing. We propose a general methodology where we sample a representative subset of UD treebanks for parser development and evaluation. Our parser uses recurrent neural networks which construct information sequentially, and we study the incorporation of a recursive neural network layer in our parser. This follows the intuition that language is hierarchical. This layer turns out to be superfluous in our parser and we study its interaction with other parts of the network. We subsequently study transitivity and agreement information learned by our parser for auxiliary verb constructions (AVCs). We suggest that a parser should learn similar information about AVCs as it learns for finite main verbs. This is motivated by work in theoretical dependency grammar. Our parser learns different information about these two if we do not augment it with a recursive layer, but similar information if we do, indicating that there may be benefits from using that layer and we may not yet have found the best way to incorporate it in our parser. We finally investigate polyglot parsing. Training one model for multiple related languages leads to substantial improvements in parsing accuracy over a monolingual baseline. We also study different parameter sharing strategies for related and unrelated languages. Sharing parameters that partially abstract away from word order appears to be beneficial in both cases but sharing parameters that represent words and characters is more beneficial for related than unrelated languages.
|
|
| 5. |
- Dubremetz, Marie, 1988-
(author)
-
Detecting Rhetorical Figures Based on Repetition of Words: Chiasmus, Epanaphora, Epiphora
- 2017
-
Doctoral thesis (other academic/artistic)abstract
- This thesis deals with the detection of three rhetorical figures based on repetition of words: chiasmus (“Fair is foul, and foul is fair.”), epanaphora (“Poor old European Commission! Poor old European Council.”) and epiphora (“This house is mine. This car is mine. You are mine.”). For a computer, locating all repetitions of words is trivial, but locating just those repetitions that achieve a rhetorical effect is not. How can we make this distinction automatically? First, we propose a new definition of the problem. We observe that rhetorical figures are a graded phenomenon, with universally accepted prototypical cases, equally clear non-cases, and a broad range of borderline cases in between. This makes it natural to view the problem as a ranking task rather than a binary detection task. We therefore design a model for ranking candidate repetitions in terms of decreasing likelihood of having a rhetorical effect, which allows potential users to decide for themselves where to draw the line with respect to borderline cases. Second, we address the problem of collecting annotated data to train the ranking model. Thanks to a selective method of annotation, we can reduce by three orders of magnitude the annotation work for chiasmus, and by one order of magnitude the work for epanaphora and epiphora. In this way, we prove that it is feasible to develop a system for detecting the three figures without an unsurmountable amount of human work. Finally, we propose an evaluation scheme and apply it to our models. The evaluation reveals that, even with a very incompletely annotated corpus, a system for repetitive figure detection can be trained to achieve reasonable accuracy. We investigate the impact of different linguistic features, including length, n-grams, part-of-speech tags, and syntactic roles, and find that different features are useful for different figures. We also apply the system to four different types of text: political discourse, fiction, titles of articles and novels, and quotations. Here the evaluation shows that the system is robust to shifts in genre and that the frequencies of the three rhetorical figures vary with genre.
|
|
| 6. |
- Edling, Agnes, 1974-
(author)
-
Abstraction and authority in textbooks : The textual paths towards specialized language
- 2006
-
Doctoral thesis (other academic/artistic)abstract
- During a few hours of a school day, a student might read textbook texts which are highly diversified in terms of abstraction. Abstraction is a central feature of specialized language and the transition from everyday language to specialized language is one of the most important things formal education can offer students. That transition is the focus of this thesis.This study introduces a new three-graded classification of abstraction including the levels of specificity, generalization and abstraction, based on a discussion of the concept of abstraction. The investigations performed, based on this classification, show that texts from different subject areas display distinct patterns of abstraction. The Swedish literary texts had the lowest degree of abstraction, the social science texts had an intermediate degree and the natural science texts were the most generalized and abstract. The results also show that the degree of abstraction in the textbook texts increases in later grade levels.The thesis presents a new way of analyzing shifts between levels of abstraction and their functions. Interestingly, the texts with a medium degree of abstraction, the social science texts, are the ones with the greatest variety in shifts. The functions of the shifts differ with respect to cultural domains. The shifts in the Swedish literary texts in general belong to the everyday domain while the shifts in the natural science texts belong to a specialized domain. The shifts in the social science texts had features of both domains.A secondary aim of the thesis is to develop the understanding of the relationship between author and reader in the texts. The results from my investigation of modality in the Swedish textbook texts confirm the earlier findings from English and Spanish textbooks. In comparison to other text types, textbook texts present knowledge in a more authoritative and less modalized way.From time to time, abstraction is described as a feature that hinders students accessing texts. Some researchers even suggest a removal of features of specialized language in textbook texts, in order to increase students’ understanding. However, in a society where specialized knowledge is necessary, the access to specialized texts is important. A democratic view of education and school mandates that children and adolescents have the opportunity to encounter and learn to encounter specialized language in school. In analyzing the texts special attention is paid to the relationship between the texts, the contexts of use and the student readers.
|
|
| 7. |
- Folkeryd, Jenny W., 1970-
(author)
-
Writing with an Attitude : Appraisal and student texts in the school subject of Swedish
- 2006
-
Doctoral thesis (other academic/artistic)abstract
- Learning in school is in many respects done through language. However, it has been shown that the language of school assignments is seldom explicitly discussed in school. Writing tasks are furthermore assigned without clear guidelines for how certain lexical choices make one text more powerful than another. The present study is a contribution to a linguistic and pedagogical discussion of student writing. More specifically the focus is on the use of evaluative language in texts written by students in the school subject of Swedish in grades 5, 8 and 11. The major investigations of the study have been accommodated within the theoretical framework of Appraisal. An overview is given of the language resources in the student texts for constructing emotion, judging behavior in ethical terms and valuing objects aesthetically. Another question addressed is that of how attitudinal meaning is intensified, thus creating greater or lesser degrees of positivity or negativity associated with the feelings. The results show that manifestations of attitude are found in practically all texts in the study. However, variations are noted in relation to different genres, age, proficiency level, language background and gender. A contribution of the study in relation to the theoretical framework upon which it draws is an extension of the system of Attitude as well as an identification of different patterns in the use of attitudinal resources. These patterns are furthermore discussed in relation to how students talk about their own written production in terms of text movability. Results indicate that students with a high degree of text movability also use attitudinal resources to a large extent. It is argued that applying the linguistic tool of Appraisal can facilitate a discussion of how to make one aspect of the hidden curriculum more visible, namely, how to write with an Attitude.
|
|
| 8. |
- Haddad, Rima, 1986-
(author)
-
Child bilingualism in Sweden and Lebanon : A study of Arabic-speaking 4-to-7-year-olds
- 2022
-
Doctoral thesis (other academic/artistic)abstract
- This dissertation investigates the vocabulary and narrative skills of 100 Arabic-Swedish-speaking children (aged 4–7 years) in Sweden cross-sectionally and the development of these skills (4 to 6) in a subgroup of 10 children longitudinally. Also, the vocabulary skills of 100 Arabic-speaking bilingual children (aged 4–7 years) in Lebanon are investigated cross-sectionally and compared to the Swedish cross-sectional study. Parental questionnaires were used to gather background information concerning language input and use inside and outside the home. The comprehension and production of vocabulary was assessed with the Cross-linguistic Lexical Task (CLT; Haman et al., 2015) and narrative macrostructure with the Multilingual Assessment Instrument for Narratives (MAIN; Gagarina et al., 2019). In Sweden, both Arabic and Swedish were investigated for vocabulary (language differences, age, socio-economic status (SES) and language input) and for narrative macrostructure (language differences, age and task effects). In Lebanon, Arabic vocabulary skills were explored in relation to age, SES and language input.Sweden: For both vocabulary and narrative macrostructure, development with age was not only evident in Swedish, but also in Arabic. Children scoring high on Arabic vocabulary comprehension and production were older and had parents speaking with them mostly in Arabic. Joint book reading in Arabic boosted the children’s Arabic expressive vocabulary whereas being exposed predominantly to Swedish had a negative effect. For Swedish, high scoring children were older and had an early age of onset of Swedish. Children who were mostly exposed to Arabic scored lower on Swedish vocabulary. Surprisingly, SES (parental education) did not predict any of the vocabulary scores. In line with international studies, narrative macrostructure production scores were generally low at this age for both languages, even for the oldest children, whereas narrative comprehension was generally well developed, even for the youngest children. The longitudinal study largely confirmed the results obtained in the cross-sectional study.Lebanon: Similarly to the Swedish sample, older children scored high on Arabic receptive and expressive vocabulary, children whose parents spoke with them mostly in Arabic scored high on expressive vocabulary, and no effects of SES were found. Compared to children in Sweden, children in Lebanon code-switched many more nouns.
|
|
| 9. |
- Hardmeier, Christian
(author)
-
Discourse in Statistical Machine Translation
- 2014
-
Doctoral thesis (other academic/artistic)abstract
- This thesis addresses the technical and linguistic aspects of discourse-level processing in phrase-based statistical machine translation (SMT). Connected texts can have complex text-level linguistic dependencies across sentences that must be preserved in translation. However, the models and algorithms of SMT are pervaded by locality assumptions. In a standard SMT setup, no model has more complex dependencies than an n-gram model. The popular stack decoding algorithm exploits this fact to implement efficient search with a dynamic programming technique. This is a serious technical obstacle to discourse-level modelling in SMT.From a technical viewpoint, the main contribution of our work is the development of a document-level decoder based on stochastic local search that translates a complete document as a single unit. The decoder starts with an initial translation of the document, created randomly or by running a stack decoder, and refines it with a sequence of elementary operations. After each step, the current translation is scored by a set of feature models with access to the full document context and its translation. We demonstrate the viability of this decoding approach for different document-level models.From a linguistic viewpoint, we focus on the problem of translating pronominal anaphora. After investigating the properties and challenges of the pronoun translation task both theoretically and by studying corpus data, a neural network model for cross-lingual pronoun prediction is presented. This network jointly performs anaphora resolution and pronoun prediction and is trained on bilingual corpus data only, with no need for manual coreference annotations. The network is then integrated as a feature model in the document-level SMT decoder and tested in an English–French SMT system. We show that the pronoun prediction network model more adequately represents discourse-level dependencies for less frequent pronouns than a simpler maximum entropy baseline with separate coreference resolution.By creating a framework for experimenting with discourse-level features in SMT, this work contributes to a long-term perspective that strives for more thorough modelling of complex linguistic phenomena in translation. Our results on pronoun translation shed new light on a challenging, but essential problem in machine translation that is as yet unsolved.
|
|
| 10. |
- Kulmizev, Artur
(author)
-
The Search for Syntax : Investigating the Syntactic Knowledge of Neural Language Models Through the Lens of Dependency Parsing
- 2023
-
Doctoral thesis (other academic/artistic)abstract
- Syntax — the study of the hierarchical structure of language — has long featured as a prominent research topic in the field of natural language processing (NLP). Traditionally, its role in NLP was confined towards developing parsers: supervised algorithms tasked with predicting the structure of utterances (often for use in downstream applications). More recently, however, syntax (and syntactic theory) has factored much less into the development of NLP models, and much more into their analysis. This has been particularly true with the nascent relevance of language models: semi-supervised algorithms trained to predict (or infill) strings given a provided context. In this dissertation, I describe four separate studies that seek to explore the interplay between syntactic parsers and language models upon the backdrop of dependency syntax. In the first study, I investigate the error profiles of neural transition-based and graph-based dependency parsers, showing that they are effectively homogenized when leveraging representations from pre-trained language models. Following this, I report the results of two additional studies which show that dependency tree structure can be partially decoded from the internal components of neural language models — specifically, hidden state representations and self-attention distributions. I then expand on these findings by exploring a set of additional results, which serve to highlight the influence of experimental factors, such as the choice of annotation framework or learning objective, in decoding syntactic structure from model components. In the final study, I describe efforts to quantify the overall learnability of a large set of multilingual dependency treebanks — the data upon which the previous experiments were based — and how it may be affected by factors such as annotation quality or tokenization decisions. Finally, I conclude the thesis with a conceptual analysis that relates the aforementioned studies to a broader body of work concerning the syntactic knowledge of language models.
|
|
| 11. |
- Lindgren, Josefin, 1985-
(author)
-
Developing narrative competence : Swedish, Swedish-German and Swedish-Turkish children aged 4–6
- 2018
-
Doctoral thesis (other academic/artistic)abstract
- This thesis investigates the development of oral narrative competence from age 4 to 6 in Swedish monolinguals (N=72) and in both languages of Swedish-German (N=46) and Swedish-Turkish (N=48) bilinguals growing up in Sweden. Picture-based fictional narratives were elicited with Cat/Dog and Baby Birds/Baby Goats from the Multilingual Assessment Instrument for Narratives (MAIN, Gagarina et al. 2012) and A2/B2 from the Edmonton Narrative Norms Instrument (ENNI, Schneider et al., 2005). Vocabulary, character introduction and narrative macrostructure were studied. Vocabulary production scores on Cross-linguistic lexical tasks (CLTs, Haman et al., 2015) were compared to NDW (number of different words) in narratives. Production of macrostructural components, macrostructural complexity, and answers to comprehension questions were analyzed. Effects of age and differences in performance between groups, between the bilinguals’ two languages, and between narrative tasks were investigated.Narrative comprehension was high already at age 4, but still developed substantially with age. In contrast, macrostructure in narrative production was at a rudimentary level at age 4. Even at age 6, the narratives contained few complete episodic structures. Children mainly included actions visible in the stimuli and rarely verbalized goals and other macrostructural components that required inferencing. The ability to introduce story characters appropriately developed strongly from age 4 to 6, but stimuli had a large effect on performance. Vocabulary showed most improvement from age 5 to 6. Development with age was clearer for the majority language Swedish than the minority languages German and Turkish, where individual variation was larger.In Swedish, pronounced differences were found between the bilingual groups. The Swedish-German bilinguals performed similarly to the monolinguals. On most measures, the Swedish-Turkish bilinguals performed lower than the other two groups, though precisely how much varied across measures. Generally, the Swedish-German children performed better in Swedish than in German, whereas the Swedish-Turkish children performed similarly in both languages or slightly higher in Turkish. The study shows that bilinguals’ two languages need not develop in parallel, and that results depend on the tasks and specific measures used. Bilingual groups differ from each other, and it is therefore not meaningful to compare all bilinguals to all monolinguals.
|
|
| 12. |
|
|
| 13. |
- Nilsson, Mattias
(author)
-
Computational Models of Eye Movements in Reading : A Data-Driven Approach to the Eye-Mind Link
- 2012
-
Doctoral thesis (other academic/artistic)abstract
- This thesis investigates new methods for understanding eye movement behavior in reading based on the use of eye tracking corpora and data-driven modeling. Eye movement behavior is characterized by two basic, generally unconscious, decisions: where and when to move the eyes. We explore the idea that empirical eye movement data carries rich information about the processes that guide these decisions. Two methods are investigated, each addressing a different aspect of eye movements in reading. The role of prediction in eye movement modeling is emphasized, and new evaluation methods for assessing the predictive accuracy of models are proposed. The decision of where to move the eyes is approached using standard machine learning methods. The model proposed learns where to move the eyes under different conditions associated with the words being read. Applied to new text, the model moves the eyes in ways it has learnt, showing characteristics similar to human readers. Furthermore, we propose the use of entropy to measure the similarity between observed and predicted eye movement behavior on held-out data. The main contribution is a flexible model, with few fixed parameters, that can be used to investigate decisions about where the eyes move during reading. The decision of when to move the eyes is approached using time-to-event modeling (survival analysis). The model proposed learns the timing of eye movements under different conditions associated with the words being read. Applied to new text, the model estimates the probability that a fixation survives for any given length of time. We propose an entropy-related measure to assess the probabilistic temporal predictions of the model. The main contribution is the use of Cox hazards modeling to address questions about the strength, as well as the timing, of processes that influence the decision of when to move the eyes during reading.
|
|
| 14. |
- Okati, Farideh, 1963-
(author)
-
The Vowel Systems of Five Iranian Balochi Dialects
- 2012
-
Doctoral thesis (other academic/artistic)abstract
- The vowel systems of five selected Iranian Balochi dialects are investigated in this study, which is the first work to apply empirical acoustic analysis to a large body of recorded data on the vowel inventories of different Balochi dialects spoken in Iran. The selected dialects are spoken in the five regions of Sistan (SI), Saravan (SA), Khash (KH), Iranshahr (IR), and Chabahar (CH) located in the province Sistan and Baluchestan in southeast Iran. The aim of the present fieldwork-based survey is to study how similar the vowel systems of these dialects are to the Common Balochi vowel system (i, iː, u, uː, a, aː, eː, oː), which is represented as the vowel inventory for the Balochi dialects in general, as well as how similar these dialects are to one another. The investigation shows that length is contrastive in these dialects, although the durational dif-ferences between the long and short counterparts are quite small in some dialects. The study also reveals that there are some differences between the vowel systems of these dialects and the Com-mon Balochi sound inventory. The Common Balochi short /i/ vowel is modified to short /e/ in these dialects, and a strong tendency for the long /eː/ and /oː/ to become the diphthongs ie and ue, respec-tively, is observed in some of the investigated dialects, specifically in KH, which shows heavier diphthongization than the other dialects. It is also observed, especially in SI, SA, and CH, that the short /u/ shows strong tendencies to shift towards a lower position of an [o] vowel. In SI and SA, this shift seems to be a correlate of syllable structure, with lowering occurring mostly in closed syllables. It is possible that Persian, as the dominant language in the area, has had an influence on these dialects and caused a lowering tendency among the higher vowels. The vowel systems in these dialects differ slightly from each other. Phonemically, the pairs e/eː, a/aː, u/uː, and the long vowels /iː/ and /oː/ are suggested for IR; the pairs a/aː, u/uː, the short /e/ and the long /iː/ as well as the diphthongs /ie/ and /ue/ substituted for the long /eː/ and /oː/, respectively, are suggested for KH; and finally the pairs e/eː, a/aː, o/oː, and the long vowels /iː/ and /uː/, which make a more symmetrical inventory, are suggested for the SI, SA, and CH dialects. In general, the vowels in these dialects show a range of phonetic variations. In addition, processes of fronting, which is most common in coronal contexts, and nasalization, which mostly occurs in nasal envi-ronments, are observed in the data researched.
|
|
| 15. |
- Pettersson, Eva, 1978-
(author)
-
Spelling Normalisation and Linguistic Analysis of Historical Text for Information Extraction
- 2016
-
Doctoral thesis (other academic/artistic)abstract
- Historical text constitutes a rich source of information for historians and other researchers in humanities. Many texts are however not available in an electronic format, and even if they are, there is a lack of NLP tools designed to handle historical text. In my thesis, I aim to provide a generic workflow for automatic linguistic analysis and information extraction from historical text, with spelling normalisation as a core component in the pipeline. In the spelling normalisation step, the historical input text is automatically normalised to a more modern spelling, enabling the use of existing taggers and parsers trained on modern language data in the succeeding linguistic analysis step. In the final information extraction step, certain linguistic structures are identified based on the annotation labels given by the NLP tools, and ranked in accordance with the specific information need expressed by the user.An important consideration in my implementation is that the pipeline should be applicable to different languages, time periods, genres, and information needs by simply substituting the language resources used in each module. Furthermore, the reuse of existing NLP tools developed for the modern language is crucial, considering the lack of linguistically annotated historical data combined with the high variability in historical text, making it hard to train NLP tools specifically aimed at analysing historical text.In my evaluation, I show that spelling normalisation can be a very useful technique for easy access to historical information content, even in cases where there is little (or no) annotated historical training data available. For the specific information extraction task of automatically identifying verb phrases describing work in Early Modern Swedish text, 91 out of the 100 top-ranked instances are true positives in the best setting.
|
|
| 16. |
-
Resourceful Language Technology : Festschrift in Honor of Anna Sågvall Hein
- 2008
-
Editorial collection (other academic/artistic)abstract
- As the first holder of the first chair in computational linguistics in Sweden, Anna Sågvall Hein has played a central role in the development of computational linguistics and language technology both in Sweden and on the international scene. Besides her valuable contributions to research, which include work on machine translation, syntactic parsing, grammar checking, word prediction, and corpus linguistics, she has been instrumental in establishing a national graduate school in language technology as well as an undergraduate program in language technology at Uppsala University. It is with great pleasure that we present her with this Festschrift to honor her lasting contributions to the field and to commemorate her retirement from the chair in computational linguistics at Uppsala University. The contributions to the Festschrift come from Anna’s friends and colleagues around the world and deal with many of the topics that are dear to her heart. A common theme in many of the articles, as well as in Anna’s own scientific work, is the design, development and use of adequate language technology resources, epitomized in the title Resourceful Language Technology.
|
|
| 17. |
- Saers, Markus, 1978-
(author)
-
Translation as Linear Transduction : Models and Algorithms for Efficient Learning in Statistical Machine Translation
- 2011
-
Doctoral thesis (other academic/artistic)abstract
- Automatic translation has seen tremendous progress in recent years, mainly thanks to statistical methods applied to large parallel corpora. Transductions represent a principled approach to modeling translation, but existing transduction classes are either not expressive enough to capture structural regularities between natural languages or too complex to support efficient statistical induction on a large scale. A common approach is to severely prune search over a relatively unrestricted space of transduction grammars. These restrictions are often applied at different stages in a pipeline, with the obvious drawback of committing to irrevocable decisions that should not have been made. In this thesis we will instead restrict the space of transduction grammars to a space that is less expressive, but can be efficiently searched. First, the class of linear transductions is defined and characterized. They are generated by linear transduction grammars, which represent the natural bilingual case of linear grammars, as well as the natural linear case of inversion transduction grammars (and higher order syntax-directed transduction grammars). They are recognized by zipper finite-state transducers, which are equivalent to finite-state automata with four tapes. By allowing this extra dimensionality, linear transductions can represent alignments that finite-state transductions cannot, and by keeping the mechanism free of auxiliary storage, they become much more efficient than inversion transductions. Secondly, we present an algorithm for parsing with linear transduction grammars that allows pruning. The pruning scheme imposes no restrictions a priori, but guides the search to potentially interesting parts of the search space in an informed and dynamic way. Being able to parse efficiently allows learning of stochastic linear transduction grammars through expectation maximization. All the above work would be for naught if linear transductions were too poor a reflection of the actual transduction between natural languages. We test this empirically by building systems based on the alignments imposed by the learned grammars. The conclusion is that stochastic linear inversion transduction grammars learned from observed data stand up well to the state of the art.
|
|
| 18. |
- Seraji, Mojgan
(author)
-
Morphosyntactic Corpora and Tools for Persian
- 2015
-
Doctoral thesis (other academic/artistic)abstract
- This thesis presents open source resources in the form of annotated corpora and modules for automatic morphosyntactic processing and analysis of Persian texts. More specifically, the resources consist of an improved part-of-speech tagged corpus and a dependency treebank, as well as tools for text normalization, sentence segmentation, tokenization, part-of-speech tagging, and dependency parsing for Persian.In developing these resources and tools, two key requirements are observed: compatibility and reuse. The compatibility requirement encompasses two parts. First, the tools in the pipeline should be compatible with each other in such a way that the output of one tool is compatible with the input requirements of the next. Second, the tools should be compatible with the annotated corpora and deliver the same analysis that is found in these. The reuse requirement means that all the components in the pipeline are developed by reusing resources, standard methods, and open source state-of-the-art tools. This is necessary to make the project feasible.Given these requirements, the thesis investigates two main research questions. The first is how can we develop morphologically and syntactically annotated corpora and tools while satisfying the requirements of compatibility and reuse? The approach taken is to accept the tokenization variations in the corpora to achieve robustness. The tokenization variations in Persian texts are related to the orthographic variations of writing fixed expressions, as well as various types of affixes and clitics. Since these variations are inherent properties of Persian texts, it is important that the tools in the pipeline can handle them. Therefore, they should not be trained on idealized data.The second question concerns how accurately we can perform morphological and syntactic analysis for Persian by adapting and applying existing tools to the annotated corpora. The experimental evaluation of the tools shows that the sentence segmenter and tokenizer achieve an F-score close to 100%, the tagger has an accuracy of nearly 97.5%, and the parser achieves a best labeled accuracy of over 82% (with unlabeled accuracy close to 87%).
|
|
| 19. |
- Serrander, Ulrika, 1976-
(author)
-
Bilingual lexical processing in single word production : Swedish learners of Spanish and the effects of L2 immersion
- 2011
-
Doctoral thesis (other academic/artistic)abstract
- Bilingual speakers cannot suppress activation from their dominant language while naming pictures in a foreign and less dominant language. Previous research has revealed that this cross-langauge activation is manifested through phonological facilitation, semantic interference and between language competition. However, this research is based exclusively on highly proficient bilinguals. The present study investigates cross-linguistic activation in Swedish learners of Spanish, grouped according to their length of Spanish immersion, and one of the groups is in its very inital stages of learning. Participants named pictures in Spanish in two picture-word interference experiments, one with only non-cognates, and one including cognates. This study addresses the following research questions; (1) do the two groups of participants differ significantly from one another in terms of cross-linguistic activation, (2) what does cross-language activation look like in initial stages of L2 acquisition, (3) how does cognate status affect cross-linguistic activation and does this differ between participants depending on length of immersion?The experiments show that cross-linguistic influence is dependent on length of immersion. The more immersed participants performed very similarly to what is usually the case in highly proficient bilinguals while the less immersed participants did not. The results of the less immersed participants are interpreted as manifestations of lexical processing in initial stages of L2 acquisition. Since this type of learner has never been tested before, there are no previous results to compare to. The results are discussed in relation to the large tradition of offline research which has shown that beginning learners predominantly process their L2 phonologically, and that conceptual processing is something requiring more L2 development.Furthermore, the cognate word induced longer naming latencies in all participants and it turned out that the cognate words were highly unfamiliar. Hence all participants are sensitive to word frequency effects, and this sensitive is greater in early stages of learning. Finally this study suggests that more research must be conducted to establish cross-linguistic influence between the many languages of multi-lingual subjects, even when these languages may not be present in the testing situation.
|
|
| 20. |
- Shao, Yan, 1990-
(author)
-
Segmenting and Tagging Text with Neural Networks
- 2018
-
Doctoral thesis (other academic/artistic)abstract
- Segmentation and tagging of text are important preprocessing steps for higher-level natural language processing tasks. In this thesis, we apply a sequence labelling framework based on neural networks to various segmentation and tagging tasks, including sentence segmentation, word segmentation, morpheme segmentation, joint word segmentation and part-of-speech tagging, and named entity transliteration. We apply a general neural CRF model to different tasks by designing specific tag sets. In addition, we explore effective ways of representing input characters, such as utilising concatenated n-grams and sub-character features, and use ensemble decoding to mitigate the effects of random parameter initialisation.The segmentation and tagging models are evaluated in a truly multilingual setup with more than 70 datasets. The experimental results indicate that the proposed neural CRF model is effective for segmentation and tagging in general as state-of-the-art accuracies are achieved on datasets in different languages, genres, and annotation schemes for various tasks. For word segmentation, we propose several typological factors to statistically characterise the difficulties posed by different languages and writing systems. Based on this analysis, we apply language-specific settings to the segmentation system for higher accuracy. Our system achieves substantially better results on languages that are more difficult to segment when compared to previous work. Moreover, we investigate conventionally adopted evaluation metrics for segmentation tasks. We propose that precision should be excluded and using recall alone is more adequate for sentence segmentation and word segmentation. The segmentation and tagging tools implemented along with this thesis are publicly available as experimental frameworks for future development as well as preprocessing tools for higher-level NLP tasks.
|
|
| 21. |
- Tang, Gongbo
(author)
-
Understanding Neural Machine Translation : An investigation into linguistic phenomena and attention mechanisms
- 2020
-
Doctoral thesis (other academic/artistic)abstract
- In this thesis, I explore neural machine translation (NMT) models via targeted investigation of various linguistic phenomena and thorough exploration of the internal structure of NMT models, in particular the attention mechanism. With respect to linguistic phenomena, I explore the ability of NMT models to translate ambiguous words, to learn long-range dependencies, to learn morphology, and to translate negation—linguistic phenomena that have been challenging for the older paradigm of statistical machine translation. I find that morphological inflection and negation are better modeled in encoder hidden states, while the senses of ambiguous words are better learned in decoder hidden states. Hidden states from lower layers are better at capturing aspects of form, such as morphological inflections and negation cues, while hidden states from higher layers are better at capturing semantic and relational aspects, such as word senses, negation events, and negation scope. I conclude that NMT models learn linguistic knowledge in a bottom-up manner. In the final part of the thesis, I interpret attention mechanisms in encoder-free models and character-level models. I show that attending to word embeddings directly does not make attention mechanisms more alignment-like but instead demonstrates that the attention mechanism is adaptable and more important for NMT than encoders. In character-level models, all characters attract equal attention except the final separators. Overall, the ability of NMT models to deal with the studied linguistic phenomena gets stronger with the evolution of architectures. NMT models perform well in translating frequent ambiguous words and learning long-range dependencies, but still suffer from morphological errors and the under-translation of negation. Attention mechanisms are crucial and adaptable, and there is no uniform behavior in different settings.
|
|
| 22. |
- Tang, Marc
(author)
-
A typology of classifiers and gender : From description to computation
- 2019
-
Doctoral thesis (other academic/artistic)abstract
- Categorization is one the most relevant tasks realized by humans during their life, as we consistently need to categorize the things and experience that we encounter. Such need is reflected in language via various mechanisms, the most prominent being nominal classification systems (e.g., grammatical gender such as the masculine/feminine distinction in French). Typological methods are used to investigate the underlying functions and structures of such systems, using a wide variety of cross-linguistic data to examine universality and variability. This analysis is itself a classification task, as languages are categorized and clustered according to their grammatical features. This thesis provides a cross-linguistic typological analysis of nominal classification systems and in parallel compares a number of quantitative methods that can be applied at different scales.First, this thesis provides an analysis of nominal classification systems (i.e., gender and classifiers) via the description of three languages with respectively gender, classifiers, and both. While the analysis of the first two languages are more of a descriptive nature and aligns with findings in the existing literature, the third language provides novel insights to the typology of nominal classification systems by demonstrating how classifiers and gender may co-occur in one language in terms of distribution of functions. Second, the underlying logic of nominal classification systems is commonly considered difficult to investigate, e.g., is there a consistent logic behind gender assignment in language? is it possible to explain the distribution of classifier languages of the world while taking into account geographical and genealogical effects? This thesis addresses the lack of arbitrariness of nominal classification systems at three different scales: The distribution of classifiers at the worldwide level, the presence of gender within a language family, and gender assignment at the language-internal level. The methods of random forests, phylogenetics, and word embeddings with neural networks are selected since they are respectively applicable at three different scales of research questions (worldwide, family-internal, language-internal).
|
|
| 23. |
- Tiedemann, Jörg, 1972-
(author)
-
Recycling Translations : Extraction of Lexical Data from Parallel Corpora and their Application in Natural Language Processing
- 2003
-
Doctoral thesis (other academic/artistic)abstract
- The focus of this thesis is on re-using translations in natural language processing. It involves the collection of documents and their translations in an appropriate format, the automatic extraction of translation data, and the application of the extracted data to different tasks in natural language processing.Five parallel corpora containing more than 35 million words in 60 languages have been collected within co-operative projects. All corpora are sentence aligned and parts of them have been analyzed automatically and annotated with linguistic markup.Lexical data are extracted from the corpora by means of word alignment. Two automatic word alignment systems have been developed, the Uppsala Word Aligner (UWA) and the Clue Aligner. UWA implements an iterative "knowledge-poor" word alignment approach using association measures and alignment heuristics. The Clue Aligner provides an innovative framework for the combination of statistical and linguistic resources in aligning single words and multi-word units. Both aligners have been applied to several corpora. Detailed evaluations of the alignment results have been carried out for three of them using fine-grained evaluation techniques.A corpus processing toolbox, Uplug, has been developed. It includes the implementation of UWA and is freely available for research purposes. A new version, Uplug II, includes the Clue Aligner. It can be used via an experimental web interface (UplugWeb).Lexical data extracted by the word aligners have been applied to different tasks in computational lexicography and machine translation. The use of word alignment in monolingual lexicography has been investigated in two studies. In a third study, the feasibility of using the extracted data in interactive machine translation has been demonstrated. Finally, extracted lexical data have been used for enhancing the lexical components of two machine translation systems.
|
|
| 24. |
- Täckström, Oscar, 1979-
(author)
-
Predicting Linguistic Structure with Incomplete and Cross-Lingual Supervision
- 2013
-
Doctoral thesis (other academic/artistic)abstract
- Contemporary approaches to natural language processing are predominantly based on statistical machine learning from large amounts of text, which has been manually annotated with the linguistic structure of interest. However, such complete supervision is currently only available for the world's major languages, in a limited number of domains and for a limited range of tasks. As an alternative, this dissertation considers methods for linguistic structure prediction that can make use of incomplete and cross-lingual supervision, with the prospect of making linguistic processing tools more widely available at a lower cost. An overarching theme of this work is the use of structured discriminative latent variable models for learning with indirect and ambiguous supervision; as instantiated, these models admit rich model features while retaining efficient learning and inference properties.The first contribution to this end is a latent-variable model for fine-grained sentiment analysis with coarse-grained indirect supervision. The second is a model for cross-lingual word-cluster induction and the application thereof to cross-lingual model transfer. The third is a method for adapting multi-source discriminative cross-lingual transfer models to target languages, by means of typologically informed selective parameter sharing. The fourth is an ambiguity-aware self- and ensemble-training algorithm, which is applied to target language adaptation and relexicalization of delexicalized cross-lingual transfer parsers. The fifth is a set of sequence-labeling models that combine constraints at the level of tokens and types, and an instantiation of these models for part-of-speech tagging with incomplete cross-lingual and crowdsourced supervision. In addition to these contributions, comprehensive overviews are provided of structured prediction with no or incomplete supervision, as well as of learning in the multilingual and cross-lingual settings.Through careful empirical evaluation, it is established that the proposed methods can be used to create substantially more accurate tools for linguistic processing, compared to both unsupervised methods and to recently proposed cross-lingual methods. The empirical support for this claim is particularly strong in the latter case; our models for syntactic dependency parsing and part-of-speech tagging achieve the hitherto best published results for a wide number of target languages, in the setting where no annotated training data is available in the target language.
|
|
| 25. |
- Wang, Luying, 1975-
(author)
-
Second Language Acquisition of Mandarin Aspect Markers by Native Swedish Adults
- 2012
-
Doctoral thesis (other academic/artistic)abstract
- This experimental study investigates the second language acquisition of the four Mandarin aspect markers -le, -guo, -zhe, and zai- by native Swedish university students enrolled in Chinese language courses in Sweden. The main points of inquiry are acquisition order, the Aspect Hypothesis, the Distributional Bias Hypothesis, and the Prototype Model. The study contains a cross-sectional study and a longitudinal study. Both written and spoken data are collected. The tasks in the cross-sectional study include film-retelling, picture-retelling, grammaticality judgment, fill-in-the-blank questions and comprehension. The longitudinal study includes written data produced by seven students in their tri-monthly journal. The study shows that perfective markers are produced before imperfective markers. The results of the experiments are consistent with the Aspect Hypothesis. The Distributional Bias Hypothesis can account for most of the Aspect Hypothesis but there are exceptions that indicate that other factors could also influence the acquisition process, such as L1 transfer. The Prototype Model cannot be conclusively proven. Apart from contributing to second-language acquisition theo-ries on cross-linguistic tense-aspect morphology, this study can provide empirical evidence with significant pedagogical implications for the second-language learning classroom.
|
|
| 26. |
- Wilhelmsen, Vera, 1976-
(author)
-
A Linguistic Description of Mbugwe with Focus on Tone and Verbal Morphology
- 2018
-
Doctoral thesis (other academic/artistic)abstract
- Mbugwe is an endangered Bantu language spoken in north central Tanzania. This PhD dissertation is a description of the Mbugwe language with a focus on tone and verbal morphology, based on the author’s fieldwork. This is the first detailed description of the language. Thus far, only a short grammatical sketch of Mbugwe has been available.Mbugwe is a tonal language with a rich verbal system. Therefore, the focus of this dissertation is on tone and verbal morphology. The dissertation also contains a brief description of the phonology of Mbugwe as well as a description of the nominal system. Mbugwe has 7 vowels; length is distinctive. There are 21 consonant phonemes, not counting the NC sequences, which are analysed as clusters in this study. Mbugwe has 17 noun classes and the adnominals that agree with the noun are described and exemplified.The tones of the language are presented in detail. Mbugwe has two tones, high and low, and the low tone is considered the default tone. The tone-bearing unit is the mora. High tones spread one mora to the right, and the last high tone before a low tone is upstepped. There is both lexical tone and grammatical tone in Mbugwe, and grammatical tones that occur on the verb stem in certain verb forms are described.In the chapter on verb morphology, the structures of the simple and periphrastic verbs are presented, as well as a description of the infinitive and copula verbs. The various tenses, aspects and moods of Mbugwe are then presented. There are 25 affirmative forms and 15 negative forms. For the perfective verbs, there are three past tenses and a future tense. In the imperfective, there is only one past tense, as well as a present and one future tense. Other aspects are the progressive, the habitual and the persistive. Moods that are grammaticalized in Mbugwe are the subjunctive, the imperative and the counterfactual. Verb forms that are not readily categorized as tense, aspect or mood are the consecutive, the situative and the participial. They depend on other verbs for their time reference.
|
|
| 27. |
- Öberg, Linnéa, 1987-
(author)
-
Words and non-words : Vocabulary and phonological working memory in Arabic-Swedish-speaking 4–7-year-olds with and without a diagnosis of Developmental Language Disorder
- 2020
-
Doctoral thesis (other academic/artistic)abstract
- This thesis investigates the vocabulary skills and the non-word repetition (NWR) performance of 99 typically developing (TD) 4–7-year-old Arabic-Swedish-speaking children and 11 Arabic-Swedish-speaking children with a diagnosis of Developmental Language Disorder (DLD). The children’s early language development, family backgrounds and language exposure patterns are explored through parental questionnaires, and for the DLD children also via interviews with parents, teachers and speech-language pathologists regarding their developmental history, language skills and communicative behaviour. Vocabulary comprehension and production is assessed with Cross-linguistic Lexical Tasks (CLT; Haman et al. 2015) in Arabic and Swedish. Phonological working memory is assessed with four different NWR tasks with varying item length, phonological complexity and language-likeness (Radeborg et al. 2006, Chiat 2015, Abou Melhem et al. 2011). For vocabulary, differences between the two languages (Arabic and Swedish) and differences between comprehension and production are explored, as well as effects of age, exposure and socioeconomic status (SES). For NWR, effects of age, task, item length and phonological complexity are investigated, as well as effects of vocabulary and exposure.Results: Vocabulary comprehension and production scores were found to increase with age in both Arabic and Swedish. Daily language exposure predicted comprehension and production scores in Arabic, but only production scores in Swedish. Length of exposure to Swedish was the most important predictor of Swedish vocabulary scores. SES (parental education) did not predict vocabulary scores in either language. For NWR, scores increased with age on all tasks. There were also task and item effects. Factors related to NWR performance were type of task, item length, phonological complexity and vocabulary skills.At group level, the DLD children scored below their TD peers on both vocabulary and NWR tasks. Many DLD children had particularly low vocabulary scores in their first language (Arabic), despite extensive and continuous exposure from birth. There was substantial overlap between the TD and the DLD groups on NWR performance, and not all DLD children scored low on NWR. Having a history of language delay or language difficulties in the family was more common among the DLD children than the TD children. The study underscores the importance of considering patterns of language exposure and developmental history when assessing the language skills of bilingual children with potential DLD.
|
|
| 28. |
- Öquist, Gustav, 1975-
(author)
-
Evaluating Readability on Mobile Devices
- 2006
-
Doctoral thesis (other academic/artistic)abstract
- The thesis presents findings from five readability studies performed on mobile devices. The dynamic Rapid Serial Visual Presentation (RSVP) format has been enhanced with regard to linguistic adaptation and segmentation as well as eye movement modeling. The novel formats have been evaluated against other common presentation formats including Paging, Scrolling, and Leading in latin-square balanced repeated-measurement studies with 12-16 subjects. Apart from monitoring Reading speed, Comprehension, and Task load (NASA-TLX), Eye movement tracking has been used to learn more about how the text presentation affects reading.The Page format generally offered best readability. Reading on a mobile phone decreased reading speed by 10% compared to reading on a Personal Digital Assistant (PDA), an interesting finding given that the display area of the mobile phone was 50% smaller. Scrolling, the most commonly used presentation format on mobile devices today, proved inferior to both Paging and RSVP. Leading, the most widely known dynamic format, caused very unnatural eye movements for reading. This seems to have increased task load, but not affected reading speed to a similar extent. The RSVP format displaying one word at time was found to reduce eye movements significantly, but contrary to common claims, this resulted in decreased reading speed and increased task load. In the last study, Predictive Text Presentation (PTP) was introduced. The format is based on RSVP and combines linguistic chunking and adaptation with eye movement modeling to achieve a reading experience that can rival traditional text presentation.It is explained why readability on mobile devices is important, how it may be evaluated in an efficient and yet reliable manner, and PTP is pinpointed as the format with greatest potential for improvement. The methodology used in the evaluations and the shortcomings of the studies are discussed. Finally, a hyper-graeco-latin-square experimental design is proposed for future evaluations.
|
|
