abstract: Transfer learning is a sub-field of statistical modeling and machine learning. It refers to methods that integrate the knowledge of other domains (called source domains) and the data of the target domain in a mathematically rigorous and intelligent way, to develop a better model for the target domain than a model using the data of the target domain alone. While transfer learning is a promising approach in various application domains, my dissertation research focuses on the particular application in health care, including telemonitoring of Parkinson’s Disease (PD) and radiomics for glioblastoma. The first topic is a Mixed Effects Transfer Learning (METL) model that can flexibly incorporate mixed effects and a general-form covariance matrix to better account for similarity and heterogeneity across subjects. I further develop computationally efficient procedures to handle unknown parameters and large covariance structures. Domain relations, such as domain similarity and domain covariance structure, are automatically quantified in the estimation steps. I demonstrate METL in an application of smartphone-based telemonitoring of PD. The second topic focuses on an MRI-based transfer learning algorithm for non-invasive surgical guidance of glioblastoma patients. Limited biopsy samples per patient create a challenge to build a patient-specific model for glioblastoma. A transfer learning framework helps to leverage other patient’s knowledge for building a better predictive model. When modeling a target patient, not every patient’s information is helpful. Deciding the subset of other patients from which to transfer information to the modeling of the target patient is an important task to build an accurate predictive model. I define the subset of “transferrable” patients as those who have a positive rCBV-cell density correlation, because a positive correlation is confirmed by imaging theory and the its respective literature. The last topic is a Privacy-Preserving Positive Transfer Learning (P3TL) model. Although negative transfer has been recognized as an important issue by the transfer learning research community, there is a lack of theoretical studies in evaluating the risk of negative transfer for a transfer learning method and identifying what causes the negative transfer. My work addresses this issue. Driven by the theoretical insights, I extend Bayesian Parameter Transfer (BPT) to a new method, i.e., P3TL. The unique features of P3TL include intelligent selection of patients to transfer in order to avoid negative transfer and maintain patient privacy. These features make P3TL an excellent model for telemonitoring of PD using an At-Home Testing Device. / Dissertation/Thesis / Doctoral Dissertation Industrial Engineering 2018

The general field of interest for this study is the Swedish public’s level of awareness of non-literal and figurative language, such as idioms and metaphors, in English. The English language can be considered to be evolving into Sweden’s second language, since it is present in so much more than school, for example, on television, computers and radio. This study aims to gain an insight into the degree of understanding of common idiomatic expressions in popular English song lyrics by native Swedish speakers. University students were chosen as the sample group and twelve of them were asked to explain their understanding of selected English idioms which occur in songs with which they were likely to recognize. The results have then been analyzed and it was possible to establish the degrees of understanding, misunderstanding and failure to understand. Firstly, the participants have explained the expression out of context, and secondly they were asked to explain it in the context of the song. This was to see whether the expressions appear to have been easier to understand out of context or in context. Earlier research show difficulties in defining the term idiom, and in this survey the term idiomatic expression has been used and defined as a way of speaking or writing that does not make any literal sense, even though the expressions are used all the time and people familiar with the language generally recognize them and understand their meaning. Earlier research has facilitated the analysis by the use of terms, theories and definitions on various ways of understanding metaphors, metonymy and idioms, and how these might, or might not, differ between Swedish and English. The results of this survey show that most of the expressions appear to have been easier to explain and understand when seen and heard in the context of the song. Although, some expressions appear to have been very alien to the participants, which has resulted in misunderstandings and failure to understand both in and out of context of the song lyrics in question.

論文提要內容： 本研究旨在探討分析台灣國中生使役動詞的使用情況，主要探討學生使用使役動詞的錯誤情況，分析錯誤原因及調查在不同英文能力的學生使用上是否存在顯著差異。 此研究採取質及量的研究。質的研究包含(1)使役動詞的補語用法(2)使役動詞的意義及用法(3)學生在使用使役動詞上的錯誤分析。量的研究包含(1)使役動詞補語的錯誤率(2)使役動詞意義及使用上的錯誤率(3)高中低程度三組中顯著變化情形。 結果顯示學生在使役動詞補語使用上常用錯非限定動詞及誤用限定動詞，此外，本研究也發現學生在選擇正確使役動詞上錯誤率高於使役動詞補語使用。分析原因如下：(1)過度規則化(2)負向轉移(3)教學引導(4)學生語言知識不足(5)語言同化。此研究也發現高中低三組在使用使役動詞補語上確實有顯著差異，然而在使役意義上(coercive sense of make-causative)、未提出但已知接受動作者意義上(unmentioned perceptible causee sense of make-causative) 、 及服務框架意義上(service frame sense of have-causative) 高中低程度學生的表現並無不同。 / Abstract The purpose of this study aims to analyze the use of periphrastic causative verbs in Junior high students’ writing performance and to understand students’ difficulty in learning and using periphrastic causative verbs. The form of periphrastic causative verbs and grammatical characteristics are easy to remember. However, subtle meanings and usage are somewhat difficult. This study adopts qualitative and quantitative analyses. The qualitative analysis includes: (1) the verb complement form of periphrastic causative verbs, (2) the meaning and functions of periphrastic causative verbs, and (3) the factors that influence students’ erroneous uses of periphrastic causative verbs. The quantitative analysis comprises: (1) the inaccuracy rate of the use of every periphrastic causative construction in the tests for form and meaning, (2) the distribution of periphrastic causative verb errors in form and meaning and use and (3) multiple comparisons among accuracy at the three proficiency levels. The results of this study are summarized as follows: students have trouble distinguishing finite verb and non-finite complement verbs with respect to the form errors. They also have lower awareness of the differences of each causative verb concerning the meaning errors. The factors affecting their incorrect responses are (1) overgeneralization, (2) negative L1 transfer, (3) teaching induced errors, (4) learners’ ignorance of rule restrictions and (5) learners’ strategies of communication and assimilation. In addition, there are significant differences among the three groups with regards to the make +O +V, have +O +V, let +O +V and have +O +p.p. in the form test and the non-coercive sense of make-causative, the hierarchical relation of have-causative and the permissive sense of let-causative in the meaning and use tests. However, there is no significant difference between the three groups’ scores of meaning and use with the coercive sense of make-causative, the unmentioned causee sense of make-causative and the service frame sense of have-causative.

Grammar is an important aspect of language learning and is thus of concern in language teaching as well. This study sets out to investigate the kinds of grammatical mistakes Swedish secondary school pupils make in their written production of English and the possible causes behind these different mistakes. Forty essays from pupils in year nine have been used to study the phenomenon of frequently occurring mistakes and their causes, and the salient mistakes have been quantatively and qualitatively accounted for in the study.      Common mistakes found in the material were subject-verb concord mistakes, mistakes involving plural nouns, incorrect article usage as well as problems using prepositions correctly. Interlingual transfer, i.e. negative transfer from the pupils’ L1, proved to be one significant factor causing mistakes in the material. However, other causes could be traced as well, such as intralingual transfer for example.      Consequently, the findings of this study point in the direction of areas which need specific attention in English L2 learning in Swedish classrooms. Thus, the study contributes with valuable insights for English teacher and teacher trainees to consider in their future practice.

This study explored the phenomenon of negative transfer from English in the writing of proficient heritage vs. second language learners of Russian. Although a number of studies have examined negative transfer from English into Russian, and other studies have compared the performance of heritage language learners (HLLs) to second language learners (SLLs), little research has investigated the phenomenon of negative transfer in the writing of both HLLs and SLLs and compared the two groups. Thus, this study employed an exploratory approach to empirically investigate the differences in language transfer between the two groups of learners. Specifically, the study addressed the following questions: (1) What are the most common types of negative transfer from English that occur in the writing of students of Russian at the Intermediate High - Advanced levels? (2) Are there differences in types and amount of negative transfer used by HLLs vs. SLLs at comparable levels of proficiency? If so, what are the differences? The research was conducted at the Russian Flagship Program at Portland State University. Participants were 13 advanced students of Russian - 6 heritage speakers of Russian and 7 native speakers of English. The study analyzed the written work of students completed during one quarter. Errors were coded into 13 categories that emerged from the data and frequencies were compared between the groups. The results found that none of the errors were more frequent than 5 per 1000 words. The most common categories were semantic extension, commas after an introductory phrase, lexical borrowings, and null subject errors. However, some categories of errors were associated with only one group, SLL or HLL, and some were not. Specifically, the categories of null subject errors, loan translations, conjunction "yesli/li" (if/whether), adjectives for nationality, and negation errors were found only in the writing of second language learners. There were only two categories limited to the group of heritage language learners: conceptual shift and number errors. The categories of semantic extension, comma after an introductory phrase, lexical borrowings, reflexive possessive pronoun "svoj", capitalization, and preposition errors were found in writing of both groups of students. Overall, second language learners made more transfer errors that heritage learners. Ultimately, this study will be of interest to anyone interested in heritage and second language acquisition of Russian, language transfer in learners of Russian, or the teaching of advanced language courses.

The ways in which someone’s first language (L1) influences his or her second language (L2) to create errors, that is negative transfer, is a topic that has received much attention in the field of Second Language Acquisition (SLA). Previous research has suggested that negative transfer is responsible for many errors. The primary aim of this study is to investigate article and preposition errors in the production of Swedish learners of English and to discuss these errors in relation to negative transfer. The secondary aim is to compare transfer errors by L1 Swedish and L1 Portuguese learners of English.   The first and main part of the study is a corpus investigation of the written production of 80 students in upper secondary school (high school). The second part is a multiple-choice test constructed to provoke transfer errors. It was distributed to students in upper secondary school in Sweden and in Brazil. The results from the corpus analysis are largely in line with those of previous research, for example with regards to how definite article errors are more common than indefinite article errors, and how contexts with definite articles and generic noun phrases seem prone to create transfer errors. The corpus study also shows that substitution was the most common preposition error and that many transfer preposition errors supposedly were caused by direct translations. Through the multiple-choice test, the degree to which the first language had an impact on individual errors could be revealed. All in all, the study reveals several aspects of negative transfer that perhaps a single-language investigation could not, because it is in the comparison of English-learners with different L1s that the most interesting results occur. / De sätt som någons första språk påverkar hans eller hennes andra språk så att fel uppstår, det vill säga negativ språköverföring (negative transfer), är ett ämne som har fått mycket uppmärksamhet inom forskning kring språkinlärning. Tidigare studier har visat på hur negativ språköverföring orsakar många fel. Det primära syftet i denna uppsats är att undersöka artikel- och prepositionsfel i skrift hos svenska elever som lär sig engelska och att diskutera dessa fel i relation till negativ språköverföring. Det sekundära målet är att jämföra hur negativ språköverföring yttrar sig hos svenska och portugisiska elever som lär sig engelska. Den första delen av denna uppsats, som även är huvuddelen, är en korpusundersökning av 80 uppsatser skrivna av högstadieelever. Den andra delen är ett flervalstest som konstruerats för att provocera fram språköverföringsfel. Deltagarna var högstadie- och gymnasieelever från Sverige och Brasilien. Resultaten från korpusundersökningen stämmer i stor utsträckning överens med vad tidigare forskning visat, till exempel att det är vanligare att fel uppstår med bestämda artiklar än med obestämda artiklar och hur bestämda artiklar och tillsammans med generiska nominalfraser tenderar att generera negativ språköverföring. Korpusstudien visar också på hur felaktigt utbyte (substitution) var det vanligaste prepositionsfelet och att många språköverföringsfel förmodligen orsakats av direkta översättningar från svenska. I analysen av resultateten från flervalstestet kunde graden av inflytande från första språket på enskilda fel påvisas. Sammantaget avslöjar undersökningen i denna uppsats på flera aspekter av negativ språköverföring som troligen ej uppenbarats om enkom elever med ett förstaspråk inkluderats, detta eftersom det är i jämförelsen mellan engelska-elever med olika förstaspråk som de mest intressanta resultaten framkommer.

This thesis deals with Ukrainian-Czech interference at the syntactical level. The theoretical section deals with the terms we work with, defines both contrastive and error analyses, and focuses on syntactic differences between Ukrainian and Czech. The central part of this section focuses on different valencies for Czech and Ukrainian verbs. The practical section consist of an experiment in which we observe the error rate of Ukrainian students when using Czech verbs that have a different valency in Ukrainian. It also includes the differential valency dictionary of Czech and Ukrainian verbs, which is the result of this thesis. Key words: interference, positive transfer, negative transfer, contrastive analysis, error analysis, valency, verb pattern, syntax, Czech, Ukrainian, vocabulary

The thesis deals with second language acquisition (SLA) of Czech in Russian and English students. It presents the main theories of SLA (generative and cognitive approaches) and compares them to the results of author's research into L2 acquisition of Czech morphology and syntax in speakers of two typologically and structurally different mothertongues. It shows that language transfer plays an important role in SLA and that some of the generative assumptions about SLA that are claimed to apply universally cannot be proven for Czech.