Cette thèse traite de l'apprentissage supervisé de transformations d'arbres XML. Le langage XML permet de décrire des données sous forme d'arbres dont la structure est définie par un schéma. Il est par conséquent devenu le standard en termes d'échanges de données, que ce soit sur le Web ou entre plusieurs applications. Toutefois, les documents XML peuvent avoir des structures très variables. La grande variété de ces structures nécessite alors d'être capable de transformer de tels arbres. Nous proposons d'effectuer de telles transformations d'arbres XML en annotant les arbres d'entrée, c'est-à-dire en associant un label à chacun de ses noeuds, la sémantique associée aux labels permettant de transformer l'arbre.<br />Afin d'apprendre à effectuer ces transformations, nous adaptons donc dans un premier temps au cas des arbres XML le modèle des champs aléatoires conditionnels ou Conditional Random Fields (CRF). Les CRFs sont un modèle graphique non dirigé conditionnel pour l'annotation : ils modélisent la probabilité conditionnelle d'une annotation sachant une observation. Ils ont, jusqu'à présent, été essentiellement utilisés dans le cadre de tâches d'annotation de séquences, à la fois dans le domaine de l'extraction d'informations ou en traitement automatiques des langues naturelles. Notre adaptation des CRFs au cas de l'annotation d'arbres XML porte à la fois sur le modèle de dépendances et sur les algorithmes d'inférence exacte (recherche de la meilleure annotation) et d'apprentissage.<br />De plus, nous proposons deux méthodes d'amélioration de la complexité de ces algorithmes afin de permettre l'utilisation des champs aléatoires conditionnels dans le cadre d'applications à grande échelle. Ces méthodes s'appuient toutes deux sur l'utilisation des connaissances du domaine. La première consiste en l'intégration de contraintes sur l'annotation. Celles-ci viennent restreindre l'espace des annotations possibles d'un arbre en interdisant des configurations de labels. La seconde technique d'amélioration de la complexité que nous proposons consiste en l'approximation d'un CRF par la composition de plusieurs CRFs de complexité moindre, définis sur des sous-parties de l'alphabet des labels.<br />Ces travaux ont été validés par diverses expériences sur des données artificielles et réelles, montrant ainsi non seulement la qualité des transformations effectuées à l'aide de nos méthodes, mais aussi leur intérêt dans des tâches réelles. Ces bons résultats nous ont conduit à réaliser une application de génération automatique de flux RSS à partir de pages Web. Celle-ci permet à son utilisateur d'apprendre un générateur de flux RSS en annotant une ou plusieurs pages d'un site Web. Ce générateur consiste en un CRF qui annote les pages Web de ce site de façon à transformer l'arbre XHTML en un arbre XML au format RSS. Il permet alors de créer automatiquement des flux RSS pour toute autre page du même site. Cette application est disponible à l'adresse suivante : http://r2s2.futurs.inria.fr/

Social conflict is a part of everyday life, and it can be a potent stressor for both humans and other animals. In the laboratory, when two Syrian hamsters (Mesocricetus auratus) compete for territory, a dominance hierarchy is quickly formed. Becoming subordinate is a significant stressor resulting in increased release of adrenocorticotropic hormone, β-endorphin, and cortisol. Defeated hamsters will also subsequently fail to display territorial aggression in future social encounters and will instead display increased submissive behavior, even in the presence of a smaller, non-aggressive intruder. This change in behavior is consistent and long-lasting and has been termed conditioned defeat (CD). Corticotropin releasing factor (CRF) is an important neuropeptide in the control of the hypothalamo-pituitary-adrenal (HPA) axis response to stress. It is also involved in a number of behaviors such as anxiety, stress responding, food intake, learning, and memory. The widespread distribution of CRF, CRF-like peptides, and CRF receptors, particularly in brain sites related to anxiety, fear, and stress responses, suggests a role for CRF and CRF-like peptides in modulating emotional responses other than via HPA axis activity. It has also been shown that CRF may have a role in the acquisition and expression of CD. Non-specific and CRF type 2-specific CRF antagonists reduce the acquisition and expression of CD in male hamsters while injection of a CRF type 1-specific antagonist does not. Therefore, the goal of this dissertation was to investigate the role of CRF type 1 and 2 receptors in CD in hamsters and to identify neuroanatomical locations where CRF may be acting. It was found that non-specific or CRF type 1 receptor specific agonists enhance the expression, but not acquisition, of CD. Further, these agonists appear to enhance aggressive behavior in animals that were not previously defeated, suggesting a modulatory role for CRF type 1 receptors in agonistic behavior that depends on an animal’s previous social experience. Further, localization of CRF receptors was determined in hamster brain in sites thought important for CD and agonistic behavior, but changes in receptor binding following defeat were not observed. Implications of these results and future directions are discussed.

The Severity of Renal Disease Scale (SORDS) was developed to provide a single score reflecting disease severity of renal patients independent of confounding psychosocial influences. This study examined SORDS' reliability and validity and its relevance as a research tool assessing the psychological effect of illness severity. Data was collected from 127 renal patients (predialysis, HD, CAPD). SORDS was compared with the Endstage Renal Disease Severity Index (ESRD-SI), the SF-36, the Beck Depression Inventory - 2nd Edition and a subset of BDI-II items reflecting cognitive features only at differing stages of renal disease and time on dialysis. SORDS and ESRD-SI data from twenty-two CAPD patients was included in reliability analyses. SORDS reliability estimates were low suggesting that the use of SORDS with medical chart data at this time is problematic. SORDS should be used only by medical practitioners who are aware of patients' standing on SORDS variables. There was however strong support for SORDS' validity. Validity was demonstrated by correlations between SORDS and the ESRD-SI. Compared to the ESRD-SI, SORDS was better able to discriminate between dialysis and pre-dialysis patients. SORDS and ESRD-SI scores were related to self-perceptions of decreased health status on the SF-36 independent of dialysis duration and age. SORDS utility in psychosocial research with renal patients was demonstrated by a finding that disease severity differentially impacts levels of depression for HD versus CAPD patients independent of age or dialysis duration. At the lowest level of illness severity as assessed by SORDS, CAPD patients scored in the moderate range of depression and were significantly more depressed than HD patients. Using the same analyses but with the ESRD-SI, no differences in level of depression were detected. These results imply a relationship between adjustment to treatment and illness severity. It is concluded that SORDS is a valid index of renal disease severity and that illness severity as assessed by SORDS may have an important role as a moderator variable in psychosocial research with renal patients. These results may have important implications for treatment assignment and psychosocial assessment and intervention of renal patients and their families.

Submitted by PMBqBM null (pmbqbm@ufba.br) on 2017-05-24T11:28:43Z No. of bitstreams: 1 Dissertação Juliana Sacramento Mota de Souza_PMBqBM-UFBA.pdf: 33818819 bytes, checksum: 7904ea127d9fe97a4dca2dc89cca3c2f (MD5) / Approved for entry into archive by Delba Rosa (delba@ufba.br) on 2017-07-06T13:24:46Z (GMT) No. of bitstreams: 1 Dissertação Juliana Sacramento Mota de Souza_PMBqBM-UFBA.pdf: 33818819 bytes, checksum: 7904ea127d9fe97a4dca2dc89cca3c2f (MD5) / Made available in DSpace on 2017-07-06T13:24:46Z (GMT). No. of bitstreams: 1 Dissertação Juliana Sacramento Mota de Souza_PMBqBM-UFBA.pdf: 33818819 bytes, checksum: 7904ea127d9fe97a4dca2dc89cca3c2f (MD5) / FAPESB / O Vírus da imunodeficiência humana do tipo 1 (HIV-1) possui uma ampla variabilidade genética. Esta diversidade é representada pelos quatro grupos, pelos nove subgrupos do grupo M e também pelas formas recombinantes destes vírus. Dentre estes, os recombinantes BF tem se destacado pela alta dispersão global concomitante com o aumento em número e diversidade. Atualmente quatorze formas recombinantes circulantes (CRFs) BF e inúmeras formas recombinantes únicas BF (URFs) foram descritas. O objetivo deste trabalho foi realizar a caracterização molecular e geográfica de isolados virais recombinantes BF do HIV- 1. Para tal, sequências e informações de genomas completos de recombinantes BF do HIV foram coletadas dos bancos de dados NCBI e Los Alamos. Posteriormente foram realizadas a caracterização destas sequências, análises filogenéticas, de recombinação e da alça V3 da proteína glicoproteína 120 (gp120), além de busca por assinaturas nucleotídicas e predição do coreceptor de entrada viral. Foram encontradas 252 sequências de genoma completo (>7000 pb) de recombinantes BF do HIV oriundas de treze países diferentes, sendo a maioria delas provenientes de pacientes brasileiros (52,8%). Seis sequências previamente caraterizadas como recombinantes BF, foram reclassificadas como subtipo B puro. Algumas sequências formaram agrupamentos monofiléticos com as CRFs já descritas e, portanto, sugere-se a reclassificação das mesmas. Dentre as sequências avaliadas foram encontrados 114 padrões distintos de recombinação. A maioria das sequências apresenta recombinação na região pol (81%; 205/252), seguido das regiões gag (68%; 172/252) e env (54%; 137/252) sendo encontrado na região tat a menor taxa de recombinação (7%; 17/252). Enquanto em todos os outros genes acessórios e regulatórios predominou o genótipo B, em vif, o subtipo F prevaleceu (38%; 96/252). Além disso, foram identificados dois hotspots entre as sequências analisadas: um encontrado em 35 sequências (13,9%) na região entre as posições 5360 – 5390 do gene acessório vif e outro em torno da posição 9356 da região nef (11% das sequências). Foram identificadas assinaturas nucleotídicas conservadas exclusivas para cada CRF_BF descrita. As mutações mais frequentes foram encontradas na região da transcriptase reversa foram M41L (NRTIs) e K100N (NNRTIs). De maneira geral, as mutações encontradas nestas sequências conferem resistência a diversos fármacos sendo que as maiores taxas de resistência foram encontradas para os antirretrovirais Atazanavir, Saquinavir, Nevirapina, Zidovudina, Estavudina, Didanosina e Emtricitabina. A predição de uso de coreceptor viral mostrou estes recombinantes utilizam preferencialmente o coreceptor CCR5 (67,5%) sendo que a CRF42 utiliza exclusivamente este coreceptor, enquanto todas as sequências da CRF39 foram classificadas como R5X4/X4. Os tetrapeptídeos mais frequentes encontrados na alça V3 foram GPGR (49,4%), GPGQ (20%), APGR (6,7%), GPGK (5,1%), GWGR (4,3%) e GPGG (2%). Além disso, as sequências com o motivo GPGQ e APGR são preferencialmente classificadas como R5 e contem o subtipo F na região da alça V3. Pode-se observar que há uma grande diversidade dos padrões de recombinação evidenciando que a recombinação entre os subtipos B e F é frequente. / Human immunodeficiency virus type 1 (HIV-1) has a wide genetic variability. This diversity is represented by the four groups, nine subgroups of group M and by the recombinant forms of these viruses. Among these, the recombinant BF has been distinguished by the high global dispersion concomitant with the increase in number and diversity. Currently fourteen BF Circulating Recombinant Forms (CRFs) and BF Unique Recombinant Forms (URFs) forms have been described. The objective of this work was to perform molecular and geographic characterization of HIV-1 recombinant BF viral isolates. To that end, complete genome sequences and information from recombinant HIV BFs were collected from the NCBI and Los Alamos databases. Subsequently, the characterization of these sequences, phylogenetic analysis, recombination and the V3 loop of the glycoprotein 120 protein (gp120) were carried out, as well as the search for nucleotide signatures and prediction of the viral entry coreceptor. A total of 252 complete genome sequences (> 7000 bp) of recombinant HIV BF from thirteen different countries were found, most of them coming from Brazilian patients (52.8%). Six sequences previously characterized as recombinant BF were reclassified as pure B subtype. Some sequences formed monophyletic clusters with the CRFs already described and, therefore, it is suggested to reclassify them. Among the sequences evaluated, 114 distinct patterns of recombination were found. Most of the sequences present recombination in the pol region (81%, 205/252), followed by the gag (68%, 172/252) and env (54%; 137/252) regions being found in the tat region at the lowest recombination rate (7%, 17/252). While in all other accessory and regulatory genes, genotype B predominated, in vif, the F subtype was dominant (38%, 96/252). In addition, two hotspots were identified among the sequences analyzed: one found in 35 sequences (13.9%) in the region between positions 5360-5390 of the vif accessory gene and another around position 9356 of the nef region (11% of sequences). Unique conserved nucleotide signatures were identified for each CRF_BF described. The most frequent mutations were found in the reverse transcriptase region were M41L (NRTIs) and K100N (NNRTIs). In general, the mutations found in these sequences confer resistance to several drugs, and the highest resistance rates were found for the antiretrovirals Atazanavir, Saquinavir, Nevirapine, Zidovudine, Stavudine, Didanosine, and Emtricitabine. The prediction of the use of viral coreceptor showed that these recombinants preferentially use the CCR5 coreceptor (67.5%) and CRF42 exclusively uses this coreceptor, while all CRF39 sequences were classified as R5X4 / X4. The most frequent tetrapeptides found in the V3 loop were GPGR (49.4%), GPGQ (20%), APGR (6.7%), GPGK (5.1%), GWGR (4.3%) and GPGG %). In addition, the sequences with the GPGQ motif and APGR are preferably classified as R5 and contain the F subtype in the V3 loop region. It can be observed that there is a great diversity of the recombination patterns evidencing that the recombination between the subtypes B and F is frequent.

Submitted by Bruna Rodrigues (bruna92rodrigues@yahoo.com.br) on 2016-09-14T13:01:32Z No. of bitstreams: 1 TeseACC.pdf: 1221580 bytes, checksum: fd3199b489e81df5b5b589830f8ed420 (MD5) / Approved for entry into archive by Marina Freitas (marinapf@ufscar.br) on 2016-09-15T13:51:39Z (GMT) No. of bitstreams: 1 TeseACC.pdf: 1221580 bytes, checksum: fd3199b489e81df5b5b589830f8ed420 (MD5) / Approved for entry into archive by Marina Freitas (marinapf@ufscar.br) on 2016-09-15T13:51:55Z (GMT) No. of bitstreams: 1 TeseACC.pdf: 1221580 bytes, checksum: fd3199b489e81df5b5b589830f8ed420 (MD5) / Made available in DSpace on 2016-09-15T13:52:06Z (GMT). No. of bitstreams: 1 TeseACC.pdf: 1221580 bytes, checksum: fd3199b489e81df5b5b589830f8ed420 (MD5) Previous issue date: 2015-05-15 / Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) / Stressful situations are a real or potential threat for psychological or physiological integrity of an individual. The underlying neurobiological substrates involved in these processes were substancially investigated through the use of animal models of stress. In this context, a crescent number of studies have used more naturalistic animal tests, such as the social defeat test. Regarding neurobiological substrates, it is known that the amygdala plays an important role in the modulation of defensive responses. This forebrain structure has several neurotransmitters and receptors with important implications in emotional states. In this context, the neuropeptide Corticotropin Releasing Factor (CRF) and its receptors, CRF1 and CRF2, have been recently investigated as an important modulatory system of defensive reactions to aversive situations. Activation CRF mechanisms in the amygdala has been postulated as a possible neurochemical substrate underlying the emotional disorders, especially anxiety disorders, induced by stress in humans. To study anxiety-related responses induced by stressors in animals, the elevated plus maze (EPM) test has been widely used. While previous studies have emphasized the role of CRF1 receptors in modulation of anxiety in rodents exposed to the EPM, the involvement of CRF2 receptors remains unclear. Few studies, however, have investigated the effects of CRF and CRF1 and CRF2 antagonists injected directly into the amygdala on the defensive responses in mice. In addition, several studies are needed to clarify the complex relationship between CRF neurotransmission of the amygdala in the etiology of anxiety disorders related to previous exposure to stress. This study investigated the role of CRF in the amygdala upon the defense reactions evaluated in the EPM in mice previously exposed to acute social defeat. Therefore, we carried out experiments to (i) characterize the effects of acute social defeat on behavior in the EPM and on the levels of plasma corticosterone; (ii) to investigate the effects of intraamygdala microinjection of CRF, CRF1 and CRF2 antagonists on the behavior of mice in the EPM and (iii) to investigate the effects of intra-amygdala microinjections of CRF1 and CRF2 antagonists on anxiety-related behaviors of mice pre-exposed to acute social defeat. Results showed that the exposure to acute social defeat stress produces anxiogenesis at short and long terms (i.e, assessed 5 min and 10 days after stress exposure), however short-term anxiety response is variable. Stress-short term effects are accompanied by increased plasma corticosterone levels. In addition, while intra-amygdala CRF increases anxiety, local injection of CRF1 (but not CRF2) receptor antagonists produced anxiolytic-like effects, suggesting a tonic role of CRF1 in the modulation of anxiety in mice exposed to the EPM. However, it was not possible to determine what is the role of CRF neurotransmission in the responses displayed by mice pre-exposed to social defeat and submitted to EPM. / O estresse é uma ameaça real ou potencial para a integridade psicológica ou fisiológica de um indivíduo e que resulta em respostas fisiológicas e/ou comportamentais. Os conhecimentos sobre estas respostas bem como sobre os substratos neurobiológicos envolvidos nestes processos só foram possíveis com o desenvolvimento de modelos animais de estresse. Dentre os vários modelos utilizados, destaca-se o modelo de derrota social por suas características mais etológicas. Em relação aos substratos neurobiológicos, é sabido que a amídala tem um importante papel na modulação de respostas defensivas. Esta estrutura encefálica possui diversos neurotransmissores e respectivos receptores com importantes implicações em estados emocionais, dentre eles o Fator de Liberação de Corticotropina (CRF). Os mecanismos de ação do CRF se dão por sua interação com os receptores CRF1 e CRF2. A ativação destes receptores na amídala tem sido postulada como um dos possíveis substratos neuroquímicos das alterações queocorrem nos transtornos comportamentais induzidos por estresse em humanos, destacando-se os transtornos de ansiedade por serem os mais prevalentes na população. Como ferramenta de estudo desses transtornos, temos o labirinto em cruz elevado (LCE), um dos mais populares modelos animais de ansiedade. Estudos no LCE apontam que o CRF1 modula a ansiedade, enquanto o papel do CRF2 não está claro. Poucos estudos, no entanto, têm investigado os efeitos do CRF, bem como de antagonistas para CRF1 e CRF2 injetados diretamente na amídala sobre as respostas defensivas de camundongos. Além disso, ainda se fazem necessários diversos estudos para entender a complexa relação entre a neurotransmissão CRFérgica da amídala na etiologia de transtornos de ansiedade relacionados a prévia exposição ao estresse. Sendo assim, o objetivo deste estudo é investigar o papel do CRF na amídala, nas reações de defesa avaliadas no LCE em camundongos previamente expostos ao estresse de derrota social agudo. Para tanto, realizou-se experimentos para (i) caracterizar os efeitos do estresse de derrota social agudo sobre o comportamento de camundongos no LCE e sobre os níveis de corticosterona plasmática; (ii) investigar os efeitos de microinjeções intra-amídala de CRF e de antagonistas CRF1 e CRF2 sobre os comportamentos de camundongos no LCE e (iii) investigar os efeitos de microinjeções intra-amídala de antagonistas CRF1 e CRF2 em camundongos pré-expostos ao estresse de derrota social agudo e submetidos ao LCE. Os resultados obtidos demonstram que o estresse de derrota social agudo é ansiogênico a curto e longo prazos, entretanto a resposta de ansiedade a curto prazo é variável. Estes mesmos efeitos a curto prazo são acompanhados por aumento do nível de corticosterona plasmática. Ainda demonstram que o CRF na amídala é ansiogênico e que há uma modulação tônica via CRF1, já o papel do CRF2 continua indeterminado. Entretanto, não foi possível determinar o papel da neurotransmissão CRFérgica nas respostas exibidas por camundongos pré-expostos ao estresse de derrota social e submetidos ao LCE.

The pulvinar, which is located in the posterior thalamus, establishes reciprocal connections with nearly all of the visual cortical areas and is consequently in a strategic position to influence their stimulus decoding processes. Projections from the pulvinar to the primary visual cortex (V1) are thought to be modulatory, altering the response of neurons without changing their basic receptive field properties. Here, we investigate this issue by studying V1 single unit responses to sine wave gratings during the reversible inactivation of the lateral posterior nucleus (LP) - pulvinar complex in the cat. We also studied the contrast response function of V1 neurons, before and during the inactivation of the LP-pulvinar complex. No change in the preferred orientation or direction selectivity of V1 neurons was observed during pulvinar inactivation. However, for the majority of the cells tested the response amplitude to the optimal stimulus was reduced. The contrast response function of neurons was fitted with the Naka-Rushton function and analysis of the effects of pulvinar deactivation revealed a diverse set of modulations: 35% of cells had a decrease in their peak response, 11% had an increase in their C50, 6% showed modulations of the slope factor and 22% exhibited changes in more than one parameter. Our results suggest that the pulvinar modulates activity of V1 neurons in a contrast dependent manner and provides gain control at lower levels of the visual cortical hierarchy. / Le pulvinar, localisé dans le thalamus postérieur, établit des connections réciproques avec la vaste majorité des aires visuelles corticales et il est ainsi dans une position stratégique afin d’influencer les processus de décodage de celles-ci. Les projections du pulvinar au cortex visuel primaire (V1) sont considérées comme étant des projections modulatrices, qui modifieraient les réponses neuronales sans toutefois changer les propriétés de base des champs récepteurs. Dans la présente étude, nous avons étudié les réponses des neurones de V1 suite à l’inactivation réversible du complexe noyau latéral postérieur (LP)-pulvinar chez le chat. Des courbes de réponse au contraste ont été générées par la présentation de réseaux ayant plusieurs niveaux de contraste pendant l’inactivation du LP-pulvinar. Aucun changement n’a été observé concernant l’orientation préférée ou la sélectivité à la direction des neurones de V1 lors de l’inactivation du pulvinar. Néanmoins, pour la majorité des cellules testées, l’amplitude de la réponse aux stimuli optimaux a été réduite. La fonction de Naka-Rushton a été appliquée aux courbes de réponse au contraste et l’analyse des effets de l’inactivation du pulvinar a montré une panoplie d’effets modulateurs : 35% des cellules ont présenté une réduction de leur réponse maximale, 11% ont eu une augmentation de leur C50, 6% ont montré une modulation de la pente et 22% des neurones ont présenté des changements dans plus d’un paramètre. Nos résultats suggèrent que le pulvinar module l’activité des neurones de V1 d’une façon dépendante du contraste et qu’il contrôle le gain des réponses des neurones des aires primaires du cortex visuel.

The use of patient data is essential in research but it is on the other hand confidential and can only be used after acquiring approval from an Ethical Board and informed consent from the individual patient. A large amount of patient data is therefore difficult to obtain if sensitive information, such as names, id numbers and contact details, are not removed from the data, by so called de-identification. Uppsala Monitoring Centre maintains the world's larges database of individual case reports of any suspected adverse drug reaction. There exists, of today, no method for efficiently de-identifying the narrative text included in these which causes countries like the United States of America reports to exclude the narratives in the reports. The aim of this thesis is to develop and evaluate a method for automatic de-identification of case narratives in reports from the WHO Global Individual Case Safety Report Database System, VigiBase. This report compares three different models, namely Regular Expressions, used for text pattern matching, and the statistical models Support Vector Machine (SVM) and Conditional Random Fields (CRF). Performance, advantages and disadvantages are discussed as well as how identified sensitive information is handled to maintain readability of the narrative text. The models developed in this thesis are also compared to existing solutions to the de-identification problem. The 400 reports extracted from VigiBase were already well de-identified in terms of names, ID numbers and contact details, making it difficult to train statistical models on these categories. The reports did however, contain plenty of dates and ages. For these categories Regular Expression would be sufficient to achieve a good performance. To identify entities in other categories more advanced methods such as the SVM and CRF are needed and will require more data. This was prominent when applying the models on the more information rich i2b2 de-identification challenge benchmark data set where the statistical models developed in this thesis performed at a competing level with existing models in the literature.

Bakgrund: Cancerrelaterad fatigue är ett vanligt tillstånd som kan uppkomma av cancersjukdom eller cancerbehandling. Det drabbar cirka 70 till 100 procent av personer med cancer. Det råder en kunskapsbrist gällande tillståndet hos sjuksköterskor och patienter vilket kan leda till en försämrad livskvalitet hos de drabbade. Vid utebliven utredning och behandling av cancerrelaterad fatigue riskerar patienter att utveckla ett sjukdomslidande. Syfte: Syftet med studien var att beskriva patienters upplevelser av att leva med cancerrelaterad fatigue. Metod: En allmän litteraturöversikt med induktiv ansats. Studien baserades på sex vetenskapliga artiklar med kvalitativ metod och en artikel med mixad metod. Samtliga artiklar analyserades med Fribergs analysmodell. Resultat: Fyra områden av upplevelser identifierades: upplevelser av kropp och själ, sociala relationer, yrkesliv och en anpassad vardag. Patienter som levde med cancerrelaterad fatigue upplevde att de påverkades kroppsligt och själsligt. Deras sociala relationer såsom relationen till vänner, familj och partner förändrades på grund av tillståndet. Några patienter upplevde att deras yrkesliv påverkades negativt. En del patienter upplevde sig vara i behov av anpassningar i sin vardag för att kunna hantera sin cancerrelaterade fatigue. Slutsats: Det framkom att cancerrelaterad fatigue påverkade patienter negativt utifrån flera aspekter på grund av den bristfälliga energin. Konsekvenserna blev att de drabbade riskerade uppleva ett sjukdomslidande till följd av tillståndet. Med hjälp av en personcentrerad vård kan sjuksköterskor sätta in relevanta omvårdnadsåtgärder som främjar hälsa hos patienterna.

Conformité Européenne (CE) approval is a complex task for producers in Europe. The producers need to search for necessary standard documents and do the tests by themselves. CE-CHECK is a website which provides document searching service, and the company engineers want to use machine learning methods to analysis the documents and the results can improve the searching system. The first task is to construct an auto keyword extraction system to analysis the standard documents. This paper performed three different machine learning methods: Conditional Random Field (CRF), joint-layer Recurrent Neural Network (RNN), and double directional Long Short-Term Memory network (Bi-LSTM), for this task and tested their performances. CRF is a traditional probabilistic model which is widely used in sequential processing. RNN and LSTM are neural network models which show impressive performance on Natural Language processing in recent years. The result of the tests was that Bi-LSTM had the best performance: the keyword extraction recall was 76.97% while RNN was 72.99% and CRF was 70.18%. In conclusion, Bi-LSTM is the best model for this keyword extraction task, and the accuracy is high enough to provide a reliable result. The model also has good robustness that it have excellent performance on documents in different fields. Bi-LSTM model can analysis all documents in less than five minutes while manual works need months, so it saved both time and cost. The results can be used in searching system and further document analysis. / Att få Conformité Européenne (CE)-godkännande är en komplicerad process för producenter i Europa. Producenterna måste söka efter nödvändiga dokument för standarder samt utföra olika tester själva. CE-CHECK är en hemsida som erbjuder söktjänster för dokument. Företagets ingenjörer vill använda maskininlärningsmetoder för att analysera dokumenten då resultaten kan förbättra söksystemet. Den första uppgiften är att konstruera ett system som automatiskt extraherar nyckelord för att analysera dokument för standarder. Detta examensarbete använde tre olika maskininlärningsmetoder och testade deras prestanda: Conditional Random Field (CRF), joint-layer Recurrent Neural Network (RNN), samt Double directional Long Short-Term Memory network (Bi-LSTM). CRF är en traditionell probabilistisk modell som ofta används inom behandling av sekventiella data. RNN och LSTM är neurala nätverksmodeller som har visat imponerande resultat inom språkteknologi de senaste åren. Resultatet av undersökningen var att Bi-LSTM presterade bäst. Modellen lyckades extrahera 76.97% av nyckelorden medan resultatet för RNN var 72.99% och för CRF var det 70.18%. Slutsatsen blev således att Bi-LSTM är det bästa valet av modell för denna uppgift och dess exakthet är tillräckligt god för att producera pålitliga resultat. Modellen är även robust då den visar goda resultat på dokument från olika forskningsområden. Bi-LSTM kan analysera alla dokument på mindre än fem minuter medan manuellt arbete skulle kräva månader. Den minskar således både tidsåtgång och kostnad. Resultaten kan användas både i söksystem samt i vidare analys av dokument.

No description available.