Whilst there are numerous studies on young people’s engagement in academic study, the internal relationship between young people and academic study is still unclear. This thesis seeks to explain the relation of young people to their academic study, in the context of Brunei Darussalam, through analysing young people’s motive hierarchy. The research is based on the understanding that young people are faced with multiple contradicting demands from the society, which evolve with their developmental age. The contradicting demands generate conflicts for young people as they participate across the different institutional practices in their everyday lives. The research entailed a semi-participatory research approach, which emphasised young people’s lived experiences, from a first-person perspective. Eight (8) young people aged 16-18 years who are studying for their GCE A Level examinations, played roles as both trained Student Researchers, as well as participants in this research. Data were collected from focus group discussions, annotated photo albums (MyAlbum) and a ‘participant self-generated’ questionnaire (MyQuestionnaire). The focus of the data collection was on the young people’s experiences of conflicts with respect to their academic study and the different agendas in their everyday lives. Intermediary tools were developed to focus the data analysis to identify motive-orientations and their relative importance in the construct of the motive hierarchy of a young person. An initial general model of motive hierarchy was developed from this study too. It is a societal demand for young people in late adolescence to be vocational and career oriented. However this study shows the eight (8) young people are also oriented towards other objects, apart from being future oriented. They can still have a dominant motive-orientation towards intimate personal relations, which usually prevails for early adolescence. Two other motive-orientations have also emerged from this study, i.e. the societal value system and self-comfort related. These different motive-orientations of the young people contradict the societal demands and create conflicts for the young people as they participate in and across the practices. These findings are important in informing intervention programmes to improve young people’s engagement in academic study.

No description available.

L'hadronthérapie joue un rôle de plus en plus important au sein des techniques de radiothérapie grâce aux propriétés balistiques des ions et, dans le cas de ceux plus lourds que les protons, à une augmentation de l'efficacité biologique dans la région tumorale. Ces caractéristiques permettent une meilleure conformation de la dose délivrée au volume tumoral et elles permettent en particulier de traiter des tumeurs radio-résistantes. Elles conduisent cependant à une grande sensibilité du parcours des ions aux incertitudes du traitement. C'est dans ce contexte qu'a été proposée la détection de radiations secondaires émises lors des interactions nucléaires induites par les ions incidents dans le patient. La tomographie par émission de positons et la détection des rayons gamma prompts ont notamment fait l'objet d'une recherche intense ces dernières années. Le réseau de formation européen ENTERVISION, soutenu par la communauté ENLIGHT, a été crée fin 2009 pour développer ce type d'imagerie et, plus généralement, traiter les incertitudes de traitement en hadronthérapie. Le travail présenté dans ce manuscrit et intitulé ≪ Modélisation et simulation des processus physiques pour l'imagerie en ligne de l'hadronthérapie ≫ est l'un des nombreux travaux issus de ce projet. Bien que le sujet soit particulièrement large, le fil conducteur de ce travail a été une étude systématique visant in fine une implémentation d'un dispositif d'imagerie ≪ gamma prompts ≫ utilisable à la fois en faisceau de protons et d'ions carbone / Hadrontherapy is taking an increasingly important role in radiotherapy thanks to the ballistic properties of ions and, for those heavier than protons, an enhancement in the relative biological effectiveness in the tumour region. These features allow for a higher tumour conformality possible and gives the opportunity to tackle the problem of radioresistant tumours. However, they may lead to a great sensitivity of ion range to treatment uncertainties, namely to morphological changes along their path. In view of this, the detection of secondary radiations emitted after nuclear interactions between the incoming ions and the patient have been long proposed as ion range probes and, in this regard, positron emitters and prompt gammas have been the matter of intensive research. The European training network ENTERVISION, supported by the ENLIGHT community, was created in the end of 2009 in order to develop such imaging techniques and more generally to address treatment uncertainties during hadrontherapy. The present work is one of the many resulting from this project, under the subject “Modelling and simulation of physics processes for in-beam imaging in hadrontherapy”. Despite the extensive range of the topic, the purpose was always to make a systematic study towards the clinical implementation of a prompt-gamma imaging device to be used for both proton and carbon ion treatments

Les sursauts gamma (GRBs pour "Gamma-Ray Bursts" en anglais) sont de brèves bouffées très énergétiques de rayonnement de haute énergie qui sont émises sur de courtes échelles de temps (fraction de seconde à plusieurs minutes). L'émission intense des sursauts gamma à haute énergie est supposée provenir d'un trou noir de masse stellaire nouvellement formé, accompagné d'un vent collimaté (i.e. un jet) se propageant à vitesse relativiste. L'émission est observée suivant deux phases successives, la phase prompte très erratique, et la phase de rémanence, moins lumineuse. Les deux instruments embarqués sur le satellite Fermi, le "Gamma-ray Burst Monitor" (GBM) et le "Large Area Telescope" (LAT), permettent d'étudier l'émission prompte des sursauts gamma sur une grande plage d'énergie (de ~10 keV à ~100 GeV). L'objectif principal de ma thèse est l'analyse et l'interprétation des propriétés spectrales et temporelles de l'émission prompte des GRBs observés par Fermi, en particulier avec les nouvelles données du LAT (Pass 8) qui ont été rendues publiques en juin 2015.La première partie de mon travail est une analyse spectrale résolue en temps de la phase prompte du sursaut GRB 090926A avec les données du GBM et du LAT. Mes résultats confirment avec un meilleur niveau de confiance la présence d'une cassure spectrale à ~400 MeV, qui est observée en coincidence avec un pic d'émission très court. Ils révèlent que cette atténuation spectrale est présente durant toute l'émission prompte du sursaut, et que l'énergie de cassure augmente jusqu'au GeV. L'interprétation de la cassure spectrale en termes d'absorption gamma ou de courbure naturelle du spectre d'émission Compton inverse (CI) dans le régime Klein-Nishina fournit des contraintes fortes sur le facteur de Lorentz du jet. Mes résultats conduisent en outre à des rayons d'émission R ∼10^14 cm qui sont compatibles avec une origine interne de l'émission du keV au GeV au-dessus de la photosphère du jet.La seconde partie de mon travail est une exploration du modèle de chocs internes développé par des collaborateurs à l'Institut d'Astrophysique de Paris (IAP). Ce modèle simule la dynamique du jet et les processus d'émission (synchrotron et CI) d'une population d'électrons accélérés aux chocs. J'ai simulé la réponse instrumentale de Fermi à un sursaut synthétique fourni par ce code numérique, et j'ai construit une fonction paramétrique qui peut être utilisée pour ajuster le modèle aux spectres de sursauts du keV au MeV. J'ai appliqué cette fonction avec succès à un échantillon de 64 sursauts brillants détectés par le GBM. J'ai aussi confronté le modèle de l'IAP au spectre d'émission prompte de GRB 090926A. Mes résultats montrent un bon accord, et j'ai identifié quelques pistes pour les améliorer. Les spectres synthétiques sont plus larges que tous les spectres dans l'échantillon du GBM. En conséquence, je discute brièvement quelques pistes de développements théoriques qui pourraient améliorer l'accord du modèle avec les observations, ainsi que des avancées observationnelles attendues dans le futur. / Gamma-Ray Bursts (GRBs) are very energetic and brief flashes of high-energy radiations which are emitted in a short time scale (fraction of a second to several minutes). The GRB bright emission is thought to be powered by a newly formed stellar-mass black hole that is accompanied by a collimated outflow (i.e. a jet) moving at a relativistic speed. The emission is observed as two successive phases: the highly variable “prompt” phase and the late and less luminous “afterglow” phase. The two instruments on board the Fermi space telescope, the Gamma-ray Burst Monitor (GBM) and the Large Area Telescope (LAT), allow the study of GRB prompt emission over a broad energy range (from ~10 keV to ~100 GeV). In June 2015, a new set of LAT data (Pass 8) was publicly released, which were generated using improved algorithms of reconstruction and classification of gamma-ray events. The main goal of my thesis is the analysis and interpretation of the spectral and temporal properties of the prompt emission phase of the GRBs observed by Fermi, especially using LAT Pass8 data.In the first part of my work, I performed a detailed time-resolved spectral analysis of the prompt phase of GRB 090926A with GBM and LAT data. My results confirm with a greater significance the spectral break at ∼400 MeV that is observed during a fast variability pulse, and they also reveal the presence of a spectral attenuation throughout the GRB prompt emission, as well as an increase of the break energy up to the GeV domain. I interpreted the spectral break in terms of gamma-ray absorption or as a natural curvature of the inverse Compton (IC) emission in the Klein-Nishina regime. Strong constraints on the jet Lorentz factor were obtained in both scenarios. My results lead also to emission radii R ∼10^14 cm, which are consistent with an internal origin of both the keV-MeV and GeV prompt emissions above the jet photosphere.The second part of my work is an exploration of the internal shock model that has been developed by collaborators at the "Institut d'Astrophysique de Paris" (IAP). This model simulates the GRB jet dynamics and the radiations (synchrotron and IC processes) from a population of shock-accelerated electrons. I simulated the response of the Fermi instruments to the synthetic GRB spectra provided by this numerical code. From these simulations, I built a new parametric function that can be used to fit the keV-MeV spectra of GRBs with the model. I applied successfully this function to a sample of 64 GBM bright GRBs. I confronted also the IAP model to the prompt emission spectrum of GRB 090926A. I obtained a relatively good agreement and I identified a couple of solutions that may improve it. The synthetic spectra are wider than any GRB spectra in the GBM sample. I present some theoretical developments that could improve the data-model agreement in the future, and I discuss possible advances from future GRB missions as well.

La protonthérapie est une technique prometteuse pour le traitement du cancer, qui se répend de plus en plus. Le pic prononcé de son profil de dose ainsi que la longueur finie du parcours des particules rendent possible un traitement plus ciblé et permettent de mieux éviter d’endommager des tissus sains. Cependant, la précision de l’irradiation s’avère également être le risque principal lors de l’utilisation de cette technique. En effet, une erreur dans la profondeur de pénétration des particules pourrait engendrer des dégâts considérables. A l’heure actuelle, aucune méthode de contrôle n’est systématiquement utilisée pour s’assurer de la qualité du traitement. Dans ce manuscrit, une méthode indirecte de mesure de la distribution de dose, basé sur la détection de gammas prompts émis le long du parcours du faisceau, est étudiée. Deux concepts de caméra collimatée uni-dimensionnelle sont comparés à l’aune de leur utilisation potentielle : une caméra à fentes parallèles et une caméra “knife-edge”. Les deux systèmes sont optimisés par simulations de Monte Carlo et des mesures sont présentés pour valider ces simulations. La comparaison se base sur la précision avec laquelle un décalage dans la chute du profil prompt gamma peut être détecté, la résolution spatiale, le coût et la taille du système. Des recommandations sont émises pour le choix de la meilleure configuration, selon différentes exigeances. Des résultats similaires sont obtenus pour les deux concepts, atteignant une précision de environ 2 mm pour un seul point de “pencil beam” correspondant à 5e7 protons. L’étude se conclue par un tour d’horizon des pistes de recherche futures qui permettraient d’utiliser un système de détection de gammas prompts dans un contexte clinique futur. / Protontherapy is a promising technique for tumor treatment that is becoming more and more widespread. The sharply peaked profile of the dose and the finite particle range allow for very conformal treatment and better sparing of healthy tissue beyond the tumor, but he precise delivery also proves to be the biggest challenge of the technique. Errors in range are a considerable risk in proton therapy and no range monitoring method is currently systematically used for quality control. In this manuscript, an indirect method of measuring the dose distribution, via the detection of secondary prompt gamma radiation emitted along the beam path, is explored. Two different one-dimensional collimated camera concepts, a multi-parallel-slit camera and a knife-edge slit camera are compared with regards to their potential use. Both systems are optimized via Monte Carlo simulation and measurements are presented for validation. The comparison is made on the basis of the precision with which a shift in the prompt gamma profile falloff edge can be retrieved by comparison with a reference profile as well as the spatial resolution, the cost, weight and bulkiness of the system and guidelines are given for choosing the best configuration for different requirements. Similar values can be obtained for both concepts, reaching a precision for the retrieval of the falloff edge of around 2 mm for a single pencil beam spot of 5×107 protons. This study concludes with an outlook on future developments and areas of investigation with the goal of reaching clinical applicability of a prompt gamma detection system.

Prompt-gamma imaging (PGI) was proposed in the 2000s as a promising in vivo range-verification method to maintain the physical advantage of proton beams by reducing unwanted range-uncertainties. Recently, PGI with a slit camera has been successfully implemented in clinical application. Despite its high accuracy and sensitivity to range deviation being shown in several studies, the clinical benefits of PGI have not yet been investigated. Hence, to fully exploit the advantages of PGI, this thesis aims to investigate the feasibility of PGI-based range verification for the automatic classification of treatment deviations and differentiation of relevant from non-relevant changes in the treatment of head-and-neck (H&N) tumors. In the first part of this thesis, the four most common types of treatment deviations in proton therapy (PT) were investigated regarding their PGI signature and by considering clinically relevant and non-relevant scenarios. A heuristic decision tree (DT) model was iteratively developed. To gain understanding of the specific signature of the error sources, different levels of geometrical complexities were explored, from simple to complex. At the simplest level, a phantom with homogeneous density was used to distinguish range-prediction and setup errors. Next, in the intermediate complexity level, a phantom with heterogeneous density was used to inspect the additional error scenarios of anatomical changes. Finally, real patient CT scans were used to investigate the relevance of changes based on clinical constraints. In the final model, a five-step filtering approach was used during pre-processing to select reliable pencil-beam-scanning spots for range verification. In this study, five features extracted from the filtered PGI data were used to classify the treatment deviation. The model is able distinguish four introduced scenarios into six classes as follows: (1) overestimation of range prediction, (2) underestimation of range prediction, (3) setup error with larger air gap, (4) setup error with smaller air gap, (5) anatomical change, and (6) non-relevant change. To ensure the application was effective, independent patient CT datasets were used to test the model. The results yielded an excellent performance of the DT classifier, with high accuracy, sensitivity, and specificity of 96%, 100%, and 85.7%, respectively. According to these findings, this model can sensitively detect treatment deviations in PT based on simulated PGI data. In the second part of this work, an alternative approach based on machine learning (ML) was taken to automatically classify the error sources. In the first stage, the two approaches were compared, using the same features as well as the same training and test datasets. The results show that the ML approach was slightly better than the heuristic DT approach in terms of accuracy. However, the performance of both approaches was excellent for the individual scenarios. Thus, these results confirm that the PGI-based data classification with five features can be applied to detect individual sources of treatment deviation in PT. In the second stage, there was an investigation of more complex and more realistic combinations of error scenarios, which was out of the scope of the DT approach. The results demonstrated that the performance of the ML-based classifiers declined in general. Furthermore, the additional features of the PG shift did not substantially improve the performance of the classifiers. As a consequence, these findings mark important issues for future research. Potentially, usage of the spatial information from the spot-based PGI data and more complex techniques such as deep learning may improve the performance of classifiers with respect to scenarios with multiple error sources. However, regardless of this, it is recommended that these findings be confirmed and validated in simulations under measurement-like conditions or with real PG measurements of H&N patients themselves. Moreover, this classification model could eventually be tested with other body sites and entities in order to assess its compatibility and adaptation requirements. In summary, this study yielded promising results regarding the automatic classification of treatment-deviation sources and the differentiation of relevant and non-relevant changes in H&N-tumor treatment in PT with PGI data. This simulation study marks an important step towards fully automated PGI-based proton-range verification, which could contribute to closing the treatment-workflow loop of adaptive therapy by supporting clinical decision-making and, ultimately, improving clinical PT.:1 Introduction 2 Background 2.1 Proton therapy 2.1.1 Rationale for proton therapy 2.1.2 Uncertainties and their mitigation 2.2 In vivo range-verification techniques 2.2.1 Range probing 2.2.2 Proton tomography 2.2.3 Magnetic resonance imaging 2.2.4 Ionoacoustic detection 2.2.5 Treatment-activated positron-emission tomography imaging 2.2.6 Prompt-gamma based detection 3 Prompt-gamma imaging with a knife-edged slit camera 3.1 Current state-of-the-art 3.2 Prompt-gamma camera system 3.3 Data acquisition and analysis 4 Error-source classification using heuristic decision tree approach 4.1 Study design 4.1.1 Case selection 4.1.2 Investigated scenarios 4.1.3 Prompt-gamma simulation and range shift determination 4.2 Development of the model 4.2.1 First-generation model 4.2.2 Second-generation model 4.2.3 Third-generation model 4.3 Model testing 4.4 Discussion: decision-tree model 5 Error-source classification using a machine-learning approach 5.1 Machine learning for classification 5.1.1 Support-vector-machine algorithm 5.1.2 Ensemble algorithm – random forest 5.1.3 Logistic-regression algorithm 5.2 Study design 5.2.1 Case selection 5.2.2 Feature selection 5.3 Model generation 5.4 Model testing 5.5 Discussion 6 Summary/ Zusammenfassung Bibliography Appendix List of Figures List of Tables List of Abbreviations

The rapid growth of textual data in the digital age presents unique challenges in domain-specific text classification, particularly the scarcity of labeled data for many applications, due to expensive cost of manual labeling work. In this thesis, we explore the applicability of prompt-learning method, which is well-known for being suitable in few-shot scenarios and much less data-consuming, as an emerging alternative to traditional fine-tuning methods, for domain-specific text classification in the context of customer-agent interactions in the retail sector. Specifically, we implemented the entire prompt-learning pipeline for the classification task, and, our investigation encompasses various strategies of prompt-learning, including fixed-prompt language model tuning strategy and tuning-free prompting strategy, along with an examination of language model selection, few-shot sampling strategy, prompt template design, and verbalizer design. In this manner, we assessed the overall performance of the prompt-learning method in the classification task. Through a systematic evaluation, we demonstrate that with the fixed-prompt language model tuning strategy, based on relatively smaller language models (e.g. T5-base with around 220M parameters), prompt-learning can achieve competitive performance (close to 75% accuracy) even with limited labeled data (up to merely 15% of full data). And besides, with the tuning-free prompting strategy, based on a regular-size language model (e.g. FLAN-T5-large with around 770M parameters), the performance can be up to around 30% accuracy with detailed prompt templates and zero-shot setting (no extra training data involved). These results can offer valuable insights for researchers and practitioners working with domain-specific textual data, prompt-learning and few-shot / zero-shot learning. The findings of this thesis highlight the potential of prompt-learning as a practical solution for classification problems across diverse domains and set the stage for future research in this area.

No description available.

No description available.

During an investigation carried out by a police officer or a gendarme, audition reports are written, the length of which can be up to several pages. The high-level goal of this thesis is to study various automatic and reliable text summarization methods to help with this time-consuming task. One challenge comes from the specific, French and judicial data that we wish to summarize; and another challenge comes from the need for reliable and factual models. First, this thesis focuses on automatic summarization evaluation, in terms of both content (how well the summary captures essential information of the source text) and factuality (to what extent the summary only includes information from or coherent with the source text). Factuality evaluation, in particular, is of crucial interest when using LLMs for judicial purposes, because of their hallucination risks. Notably, we propose a light variation of SelfCheckGPT, which has a stronger correlation with human judgment (0.743) than the wide-spread BARTScore (0.542), or our study dataset. Other paradigms, such as Question-Answering, are studied in this thesis, which however underperform compared to these. Then, extractive summarization methods are explored and compared, including one based on graphs via the TextRank algorithm, and one based on greedy optimization. The latter (overlap rate: 0.190, semantic similarity: 0.513) clearly outperforms the base TextRank (overlap rate: 0.172, semantic similarity: 0.506). An improvement of the TextRank with a threshold mechanism is also proposed, leading to a non-negligible improvement (overlap rate: 0.180, semantic similarity: 0.513). Finally, abstractive summarization, with pre-trained LLMs based on a Transformer architecture, is studied. In particular, several general-purpose and multilingual models (Llama-2, Mistral and Mixtral) were objectively compared on a summarization dataset of judicial procedures from the French police. Results show that the performances of these models are highly related to their size: Llama-2 7B struggles to adapt to uncommon data (overlap rate: 0.083, BARTScore: -3.099), while Llama-2 13B (overlap rate: 0.159, BARTScore: -2.718) and Llama-2 70B (overlap rate: 0.191, BARTScore: -2.479) have proven quite versatile and efficient. To improve the performances of the smallest models, empirical prompt-engineering and parameter-efficient fine-tuning are explored. Notably, our fine-tuned version of Mistral 7B reaches performances comparable to those of much larger models (overlap rate: 0.185, BARTScore: -2.060), without the need for empirical prompt-engineering, and with a linguistic style closer to what is expected. / Under en utredning som görs av en polis eller en gendarm skrivs förhörsprotokoll vars längd kan vara upp till flera sidor. Målet på hög nivå med denna rapport är att studera olika automatiska och tillförlitliga textsammanfattningsmetoder för att hjälpa till med denna tidskrävande uppgift. En utmaning kommer från de specifika franska och rättsliga uppgifter som vi vill sammanfatta; och en annan utmaning kommer från behovet av pålitliga, sakliga och uppfinningsfria modeller. För det första fokuserar denna rapport på automatisk sammanfattningsutvärdering, både vad gäller innehåll (hur väl sammanfattningen fångar väsentlig information i källtexten) och fakta (i vilken utsträckning sammanfattningen endast innehåller information från eller överensstämmer med källtexten). Faktautvärdering, i synnerhet, är av avgörande intresse när man använder LLM för rättsliga ändamål, på grund av deras hallucinationsrisker. Vi föreslår särskilt en lätt variant av SelfCheckGPT, som har en starkare korrelation med mänskligt omdöme (0,743) än den utbredda BARTScore (0,542), eller vår studiedatauppsättning. Andra paradigm, såsom Question-Answering, studeras i denna rapport, som dock underpresterar jämfört med dessa. Sedan utforskas och jämförs extraktiva sammanfattningsmetoder, inklusive en baserad på grafer via TextRank-algoritmen och en baserad på girig optimering. Den senare (överlappning: 0,190, semantisk likhet: 0,513) överträffar klart basen TextRank (överlappning: 0,172, semantisk likhet: 0,506). En förbättring av TextRank med en tröskelmekanism föreslås också, vilket leder till en icke försumbar förbättring (överlappning: 0,180, semantisk likhet: 0,513). Slutligen studeras abstrakt sammanfattning, med förutbildade LLM baserade på en transformatorarkitektur. I synnerhet jämfördes flera allmänna och flerspråkiga modeller (Llama-2, Mistral och Mixtral) objektivt på en sammanfattningsdatauppsättning av rättsliga förfaranden från den franska polisen. Resultaten visar att prestandan för dessa modeller är starkt relaterade till deras storlek: Llama-2 7B kämpar för att anpassa sig till ovanliga data (överlappning: 0,083, BARTScore: -3,099), medan Llama-2 13B (överlappning: 0,159, BARTScore: -2,718) och Llama-2 70B (överlappning: 0,191, BARTScore: -2,479) har visat sig vara ganska mångsidiga och effektiva. För att förbättra prestandan för de minsta modellerna utforskas empirisk prompt-teknik och parametereffektiv finjustering. Noterbart är att vår finjusterade version av Mistral 7B når prestanda som är jämförbara med de för mycket större modeller (överlappning: 0,185, BARTScore: -2,060), utan behov av empirisk prompt-teknik och med en språklig stil som ligger närmare vad som förväntas.