<p style="text-align: justify; ">Cilj - Cilj istraživanja je razvoj modela, implementacija prototipa i verifikacija sistema za ekstrakciju metodologija iz naučnih članaka iz oblasti Informatike. Da bi se, pomoću tog sistema, naučnicima mogao obezbediti bolji uvid u metodologije u svojim oblastima potrebno je ekstrahovane metodolgije povezati sa metapodacima vezanim za publikaciju iz koje su ekstrahovani. Iz tih razloga istraživanje tako&ntilde;e za cilj ima i razvoj modela sistema za automatsku ekstrakciju metapodataka iz naučnih članaka. Metodologija - Ekstrahovane metodologije se kategorizuju u četiri kategorije: kategorizuju se u četiri semantičke kategorije: zadatak (Task), metoda (Method), resurs/osobina (Resource/Feature) i implementacija (Implementation). Sistem se sastoji od dva nivoa: prvi je automatska identifikacija metodolo&scaron;kih rečenica; drugi nivo vr&scaron;i prepoznavanje metodolo&scaron;kih fraza (segmenata). Zadatak ekstrakcije i<br />kategorizacije formalizovan je kao problem označavanja sekvenci i upotrebljena su četiri zasebna Conditional Random Fields modela koji su zasnovani na sintaktičkim frazama. Sistem je evaluiran na ručno anotiranom korpusu iz oblasti Automatske Ekstrakcije Termina koji se sastoji od 45 naučnih članaka. Sistem za automatsku ekstrakciju metapodataka zasnovan je na klasifikaciji. Klasifikacija metapodataka vr&scaron;i se u osam unapred definisanih sematičkih kategorija: Naslov, Autori, Pripadnost, Adresa, Email, Apstrakt, Ključne reči i Mesto publikacije. Izvr&scaron;eni su eksperimenti sa<br />svim standardnim modelima za klasifikaciju: naivni bayes, stablo odlučivanja, k-najbližih suseda i ma&scaron;ine potpornih vektora. Rezultati - Sistem za ekstrakciju<br />metodologija postigao je sledeće rezultate: F-mera od 53% za identifikaciju Task i Method kategorija (sa precizno&scaron;ću od 70%) dok su vrednosti za F-mere za Resource/Feature i Implementation kategorije bile 60% (sa precizno&scaron;ću od 67%) i 75% (sa precizno&scaron;ću od 85%) respektivno. Nakon izvr&scaron;enih klasifikacionih eksperimenata, za sistem za ekstrakciju metapodataka, utvr&ntilde;eno je da ma&scaron;ine potpornih vektora (SVM) pružaju najbolje performanse. Dobijeni rezultati SVM modela su generalno dobri, F-mera preko 85% kod skoro svih kategorija, a preko 90% kod većine. Ograničenja istraživanja/implikacije - Sistem za ekstrakciju metodologija, kao i sistem za esktrakciju metapodataka primenljivi su samo na naučne članke na<br />engleskom jeziku. Praktične implikacije - Predloženi modeli mogu se, pre svega, koristiti za analizu i pregled razvoja naučnih oblasti kao i za kreiranje sematički bogatijih informacionih sistema naučno-istraživačke delatnosti.<br />Originalnost/vrednost - Originalni doprinosi su sledeći: razvijen je model za ekstrakciju i semantičku kategorijzaciju metodologija iz naučnih članaka iz oblasti<br />Informatike, koji nije opisan u postojećoj literaturi. Izvr&scaron;ena je analiza uticaja različitih vrsta osobina na ekstrakciju metodolo&scaron;kih fraza. Razvijen je u potpunosti automatizovan sistem za ekstrakciju metapodataka u informacionim sistemima naučno-istraživačke delatnosti.</p> / <p style="text-align: justify; "> Purpose - The purpose of this research&nbsp;is model development, software&nbsp;prototype implementation and&nbsp;verification of the system for the&nbsp;identification of methodology mentions&nbsp;in scientific publications in a subdomain&nbsp;of automatic terminology&nbsp;extraction. In order to provide a better&nbsp;insight for scientists into the<br /> methodologies in their fields extracted&nbsp;methodologies should be connected&nbsp;with the metadata associated with the&nbsp;publication from which they are&nbsp;extracted. For this reason the purpose&nbsp;of this research was also a development&nbsp;of a system for the automatic extraction&nbsp;of metadata from scientific&nbsp;publications.&nbsp;Design/methodology/approach -&nbsp;Methodology mentions are categorized&nbsp;in four semantic categories: Task,&nbsp;Method, Resource/Feature and&nbsp;Implementation. The system comprises&nbsp;two major layers: the first layer is an&nbsp;automatic identification of&nbsp;methodological sentences; the second&nbsp;layer highlights methodological phrases&nbsp;(segments). Extraction and&nbsp;classification of the segments was&nbsp;171&nbsp;formalized as a sequence tagging&nbsp;problem and four separate phrase-based&nbsp;Conditional Random Fields were used&nbsp;to accomplish the task. The system has&nbsp;been evaluated on a manually&nbsp;annotated corpus comprising 45 full&nbsp;text articles. The system for the&nbsp;automatic extraction of metadata from&nbsp;scientific publications is based on&nbsp;classification. The metadata are&nbsp;classified eight pre-defined categories:<br /> Title, Authors, Affiliation, Address,&nbsp;Email, Abstract, Keywords and&nbsp;Publication Note. Experiments were&nbsp;performed with standard classification&nbsp;models: Decision Tree, Naive Bayes,&nbsp;K-nearest Neighbours and Support&nbsp;Vector Machines.&nbsp;Findings - The results of the system for&nbsp;methodology extraction show an Fmeasure&nbsp;of 53% for identification of&nbsp;both Task and Method mentions (with&nbsp;70% precision), whereas the Fmeasures&nbsp;for Resource/Feature and&nbsp;Implementation identification was 60%&nbsp;(with 67% precision) and 75% (with&nbsp;85% precision) respectively. As for the&nbsp;system for the automatic extraction of&nbsp;metadata Support Vector Machines&nbsp;provided the best performance. The Fmeasure&nbsp;was over 85% for almost all of&nbsp;the categories and over 90% for the&nbsp;most of them.&nbsp;Research limitations/implications -&nbsp;Both the system for the extractions of&nbsp;methodologies and the system for the&nbsp;extraction of metadata are only&nbsp;applicable to the scientific papers in&nbsp;English language.&nbsp;172&nbsp;Practical implications - The proposed&nbsp;models can be used in order to gain&nbsp;insight into a development of a&nbsp;scientific discipline and also to create&nbsp;semantically rich research activity<br /> information systems.&nbsp;Originality/Value - The main original&nbsp;contributions are: a novel model for the&nbsp;extraction of methodology mentions&nbsp;from scientific publications. The&nbsp;impact of the various types of features&nbsp;on the performance of the system was&nbsp;determined and presented. A fully&nbsp;automated system for the extraction of&nbsp;metadata for the rich research activity<br /> information systems was developed.</p>

Trabalho sem a folha de aprova??o. / Submitted by Rodrigo Martins Cruz (rodrigo.cruz@ufvjm.edu.br) on 2015-12-17T17:14:54Z No. of bitstreams: 2 pedro_ivo_sodre_amaral.pdf: 443054 bytes, checksum: 017fd5b0f5e19a7bb1100ad26a63d3ae (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Approved for entry into archive by Rodrigo Martins Cruz (rodrigo.cruz@ufvjm.edu.br) on 2015-12-17T17:15:27Z (GMT) No. of bitstreams: 2 pedro_ivo_sodre_amaral.pdf: 443054 bytes, checksum: 017fd5b0f5e19a7bb1100ad26a63d3ae (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Made available in DSpace on 2015-12-17T17:15:27Z (GMT). No. of bitstreams: 2 pedro_ivo_sodre_amaral.pdf: 443054 bytes, checksum: 017fd5b0f5e19a7bb1100ad26a63d3ae (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Previous issue date: 2012 / Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior (CAPES) / O experimento foi conduzido no munic?pio de Couto de Magalh?es de Minas/MG. Foram utilizados 36 su?nos (59,04?5,55 kg) distribu?dos em delineamento inteiramente ao acaso em tr?s tratamentos com seis repeti??es, sendo a unidade experimental composta por dois su?nos na baia. Os animais foram submetidos a tr?s programas hemerais de luz, caracterizando os tratamentos: luz natural (LN), 16 horas de luz e oito horas de escuro (16L:8E) e 23 horas de luz e uma de escuro (23L:1E). A ra??o e a ?gua foram fornecidas ? vontade, os comedouros foram repostos em tr?s tratos ?s 08h, 14h e 19h. Foi anotado o consumo de ra??o di?rio (CRD), o ganho de peso di?rio (GPD), e a convers?o alimentar (CA). Os par?metros fisiol?gicos observados foram a temperatura retal (TR), a freq??ncia respirat?ria (FR) e a temperatura superficial (TS), avaliados em dois per?odos, uma vez por semana. Os leit?es foram observados durante o per?odo nictemeral a cada dez minutos, totalizando quatro dias de avalia??o. Os padr?es observados foram: em p?, deitado, comendo, bebendo, fu?ando em p?, fu?ando outro, sentado, urinando e defecando. Para os dados de desempenho foi feita an?lise em delineamento inteiramente ao acaso, sendo a baia a parcela e o peso inicial como co-vari?vel, as m?dias foram comparadas por teste t a 5% de signific?ncia. Para as caracter?sticas fisiol?gicas (TR, FR e TS), as an?lises foram feitas em delineamento inteiramente ao acaso com esquema fatorial (3x2), considerando os dias de coleta como repeti??es, os programas de luz e os per?odos de coleta como tratamentos. As observa??es de comportamento foram submetidas ao teste de Kruskal-Wallis. As an?lises foram feitas dentro de cada hora do dia, al?m disso, avaliou-se o per?odo diurno (06h ?s 18h) e noturno (18h ?s 06h). Os programas de ilumina??o utilizados n?o influenciaram o consumo de ra??o di?rio, o ganho de peso di?rio nem a convers?o alimentar dos animais. Os resultados para TS foram maiores para os su?nos que receberam 23L:1E, no per?odo da tarde em rela??o aos demais. A luz suplementar promoveu eleva??o da FR, apenas no per?odo da tarde. A TR n?o variou entre os tratamentos. As vari?veis comportamentais ?em p??, ?defecando?, ?fu?ar em p??, ?fu?ar o outro? e ?urinando? n?o diferiram em nenhum dos per?odos estudados. Houve diferen?a significativa entre os tratamentos para a vari?vel ?deitado?, "comendo", "bebendo" e "sentado" indicando poss?vel aumento na atividade dos animais no per?odo da noite. O uso de programas de ilumina??o para su?nos em termina??o n?o alterou o desempenho dos animais. Todavia, o fornecimento de luz adicional influenciou o comportamento dos animais que ficaram mais ativos. A maior atividade dos su?nos que receberam luz apresentou, como consequ?ncia, altera??es nos par?metros fisiol?gicos indicando maior desconforto ambiental no per?odo da tarde. / Disserta??o (Mestrado) ? Programa de P?s-Gradua??o em Zootecnia, Universidade Federal dos Vales do Jequitinhonha e Mucuri, [2012]. / ABSTRACT The experiment was conducted in Couto de Magalh?es de Minas/ MG. Was used 36 pigs (59.04 ? 5.55 kg) allotted to completely randomized design with three treatments and six replicates of one experimental unit consisted of two pigs at bay. The animals underwent three light programs, featuring treatments: natural light (NL), 16 hours of light and eight hours of dark (16L: 8D) and 23 hours of light and one dark (23L: 1D). Feed and water were given ad libitum, the feeders were restored in three tracts at 08h, 14h and 19h. It was noted the daily feed intake (DFI), daily weight gain (ADG) and feed conversion (FC). The physiological parameters were observed rectal temperature (RT), respiratory rate (RR) and surface temperature (TS), evaluated in two periods, once a week. The animals were observed during the diel every ten minutes, totaling four days of evaluation. The patterns observed were: standing, lying, eating, drinking, standing digging, digging another, sitting, urinating and defecating. For performance data analysis was performed in a completely randomized design, with the pen portion and initial weight as a covariate, the means were compared by t test at 5% significance for the physiological characteristics (TR, FR and TS) analyzes were done in a completely randomized design with factorial arrangement (3x2), considering the collection days as replicates, the lighting programs and collection period as treatments. The behavioral observations were submitted to Kruskal-Wallis test. The analyzes were performed within each time of day, in addition, we assessed the daytime (06h to 18h) and night (18h to 06h). The lighting programs used did not affect feed intake, daily weight gain or feed the animals. The results for TS were higher for pigs receiving 23L: 1D, in the afternoon compared to the other. The supplemental light promoted an increase in RF only in the afternoon. The TR did not vary between treatments. The behavioral variables "standing", "defecating", "tweak in the foot", "tweak the other" and "pissing" did not differ in any of the periods studied. There were significant differences between treatments for the variable "lying", "eating", "drinking" and "sitting" indicating a possible increase in the activity of the animals at night. The use of lighting programs for finishing pigs did not affect animal performance. However the provision of additional light influenced the behavior of animals that were more active. A greater activity of the pigs that had received the light, as a consequence, changes in physiological parameters indicating greater environmental discomfort in the afternoon.

Submitted by Celso Magalhaes (celsomagalhaes@ufrrj.br) on 2017-05-19T11:27:05Z No. of bitstreams: 1 2015 - Jos? Claudio Bezerra M. Junior.pdf: 1045163 bytes, checksum: d759d4bffb13bbf9696edda30c7326f2 (MD5) / Made available in DSpace on 2017-05-19T11:27:05Z (GMT). No. of bitstreams: 1 2015 - Jos? Claudio Bezerra M. Junior.pdf: 1045163 bytes, checksum: d759d4bffb13bbf9696edda30c7326f2 (MD5) Previous issue date: 2015-12-02 / Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior - CAPES / The determination of digestible lysine requirements for Nile tilapia is essential to adjust the correct balance of amino acids in it?s food. Lysine is the amino acid reference for studies in ideal protein, which is currently an important concept in animal nutrition once it allows a reduction in the protein amount of the diet without compromising performance, and furthermore reducing the nitrogen excretion and eutrophication risks of the water. This study aimed to determine the digestible lysine requirement for Nile tilapia in the final phase of growth (500 to 600 g of body weight). Isocaloric and isoproteic experimental diets were utilized with different levels of digestible lysine in each formulation. The amounts of L-lysine HCl 78% used were 0.000; 0.388; 0.776; 1.164, and 1.552% replacing the ingredient glutamic acid 99%; which adjust the treatments to 0.932; 1.23; 1.53; 1.83, and 2.14% of digestible lysine. Nile tilapia weighing 519 g making a sum of 300 animals were utilized in this experiment. They were distributed in 25 water tanks each one measuring a 1000 liters using a renewable system of water. The pH, dissolved oxygen, temperature, salinity and conductivity were monitored daily and ammonia was measured weekly. Slaughters happened when fishes reached 28 and 50 days counting from the beginning of the experiment. The following animal science aspects were evaluated at the end: weight gain (WG), daily weight gain ratio (DWGR), specific growth rate (SGR), food intake (FI), consumption of digestible lysine (CDL), food conversion (FC) protein efficiency gain (PEG), lysine efficiency gain (LEG), nitrogen retention efficiency (NRE), protein deposition rate (PDR), Daily fat deposition rate (DFDR), protein retention efficiency (PRE), and survival rate (SR). The data was interpreted by analysis of variance with 5% probability. There was quadratic effect for GP, CA, TCE, ELG, EPG. The estimated value of lysine for both the largest GP and the best CA is 1.31% digestible lysine. There was no significant differency regarding the food intake, carcass or filet characteristics. This result is in part due to the fact of the gain being proportional throughout the body of the fish. In the second slaughter were obtained quadratic effect in the ether extract and carcass moisture. The Nile tilapia in the weight range of 500 to 600 grams of body weight presents the requirement for 1.31% of digestible lysine, corresponding to 5.31% of digestible dietary protein and 0.431% per Mcal of digestible energy for greater weight gain and better food conversion in it?s experiment conditions / A determina??o da exig?ncia de lisina para a til?pia-do-Nilo ? importante para a elabora??o de ra??es com adequado balanceamento de amino?cidos. A lisina ? o amino?cido refer?ncia nos estudos de prote?na ideal, um conceito relevante atualmente uma vez que permite a redu??o na quantidade de prote?na da ra??o sem comprometimento do desempenho, diminuindo ainda a excre??o de nitrog?nio para a ?gua e o risco de eutrofiza??o. O presente trabalho objetivou determinar a exig?ncia em lisina digest?vel para a til?pia-do-Nilo na fase de 500 a 600 g de peso vivo. Foram utilizadas dietas experimentais que consistiram em ra??es isocal?ricas, isoproteicas e com valores crescentes de lisina digest?vel. A L-lisina HCl 78% foi utilizada com valores crescentes (0,000; 0,388; 0,776; 1,164 e 1,552%), em substitui??o ao ingrediente ?cido glut?mico (99%), perfazendo os tratamentos com 0,932; 1,23; 1,53; 1,83 e 2,13% de lisina digest?vel. Foram utilizadas 300 til?pias-do-Nilo com peso m?dio de 519 g e distribu?dos em 25 caixas d??gua de 1000 L em sistema aberto de circula??o de ?gua. O pH, oxig?nio dissolvido, temperatura, salinidade e condutividade da ?gua foram monitorados diariamente e a am?nia foi medida semanalmente. Foram realizados 2 abates, aos 28 dias e aos 50 dias de experimento. Foram avaliados os seguintes ?ndices zoot?cnicos: ganho de peso (GP), ganho de peso di?rio (GPD), taxa de crescimento espec?fico (TCE), ingest?o alimentar (IA), consumo de lisina digest?vel (CLD), convers?o alimentar (CA), efici?ncia proteica para ganho (EPG), efici?ncia de lisina para ganho (ELG), efici?ncia de reten??o de nitrog?nio (ERN), taxa de deposi??o de prote?na (TDP), taxa de deposi??o di?ria de gordura (TDG), efici?ncia de reten??o de prote?na (ERP) e taxa de sobreviv?ncia (TS). Os dados foram interpretados por meio de an?lise de vari?ncia a 5% de probabilidade. Houve efeito quadr?tico para GP, CA, TCE, ELG, EPG, sendo o valor estimado de lisina tanto para o maior GP quanto para a melhor CA de 1,31% de lisina digest?vel. N?o houve diferen?a significativa para o consumo de ra??o, rendimento de carca?a, rendimento de fil?, caracter?sticas da carca?a e do fil?. Esses resultados se devem em parte ao fato do ganho ter sido proporcional em todo o corpo do peixe. No segundo abate houve efeito quadr?tico no extrato et?reo e na umidade da carca?a. A Til?pia-do-Nilo na faixa de peso de 500 a 600 gramas de peso vivo apresenta a exig?ncia de 1,31% de lisina digest?vel, que corresponde a 5,31% da prote?na digest?vel da dieta e a 0,431%/Mcal de energia digest?vel para o maior ganho de peso e melhor convers?o alimentar nas condi??es deste experimento.