Thesis (MScEng (Electrical and Electronic Engineering))--Stellenbosch University, 2008. / Automatic bird song recognition and transcription is a relatively new field. Reliable automatic recognition systems would be of great benefit to further research in ornithology and conservation, as well as commercially in the very large birdwatching subculture. This study investigated the use of Hidden Markov Models and duration modelling for bird call recognition. Through use of more accurate duration modelling, very promising results were achieved with feature vectors consisting of only pitch and volume. An accuracy of 51% was achieved for 47 calls from 39 birds, with the models typically trained from only one or two specimens. The ALS pitch tracking algorithm was adapted to bird song to extract the pitch. Bird song synthesis was employed to subjectively evaluate the features. Compounded Selfloop Duration Modelling was developed as an alternative duration modelling technique. For long durations, this technique can be more computationally efficient than Ferguson stacks. The application of approximate string matching to bird song was also briefly considered.

Interdisciplinarity is a common goal within specialized interdisciplinary organizations as well as within the academic world as a whole. Interdisciplinary research is connected with values such as a more market oriented approach, new ways of using theories and methods and a more cost efficient use of expensive research equipment. But few studies actually measure the development of organized interdisciplinary research over time. The purpose of this paper is to examine the potential effects of organizational changes in research institu- tions whose purpose is to promote interdisciplinary research collaborations.. Using a longitudinal case study this essay measures the attempts of a Norwegian music therapy research organisation to actively induce interdisciplinary research over a period of eight years. An important part of the study is also to try common quantitative operationalizations of different types of interdisciplinarity with in a small case study. Bibliometric methods are used to visualize relations between researchers. Important results include findings of increased interdisciplinary activity within the new organisation thus proving the possibility to create interdisciplinary conditions thru active organisational changes. The essay also show that operationalizations used in quantitative studies may need to be adapted to specific conditions in order to effectively measure interdisciplinary relations between researchers. As with all case studies the results of this essay needs to be tested and verified in further studies.

This study presents a theory of melodic creation, transmission, memory, and recall within the Anglo- and Celtic-American culture of lower Appalachia, from the time of the earliest European settlers until the present. This theory and its attendant hypotheses draw upon earlier published ideas, current theories of memory and recall, and the results of applying a computer-supported analytical system developed by the author. Sources include previous studies of folksong melody, song collections, and earlier investigations of the psychology of memory. Also important are portions of an anonymous treatise on traditional Celtic musical scales and an authoritative, modern interpretation of this document. A final body of sources is a small group of sound-recordings.

Large language models (LLMs) are very large deep learning models which are retrained on a huge amount of data. Among the LLMs are sentence bidirectional encoder representations from transformers (SBERT) where advanced training methods such as transformer-based denoising autoEncoder (TSDAE), generative query network (GenQ) and an adaption of generative pseudo labelling (GPL) can be applied. This thesis project aims to develop a semantic search tool for Swedish legal judgments in order to overcome the limitations of traditional keyword searches in legal document retrieval. For this aim, a model adept at understanding the semantic nuances of legal language has been developed by leveraging natural language processing (NLP) and fine- tuning LLMs like SBERT, using advanced training methods such as TSDAE, GenQ, and an adaption of GPL. To generate labeled data out of unlabelled data, a GPT3.5 model was used after it was fine-tuned. The generation of labeled data with the use of a generative model was crucial for this project to train the SBERT efficiently. The search tool has been evaluated. The evaluation demonstrates that the search tool can accurately retrieve relevant documents based on semantic queries and simnifically improve the efficiency and accuracy of legal research. GenQ has been shown to be the most efficient training method for this use case.

Thesis (MScEng)--Stellenbosch University, 2011. / ENGLISH ABSTRACT: A new type of wideband lters, with tunability in both the centre frequency and bandwidth, is presented in this thesis. These lters are based on perturbed ring-resonators in cascade, while varactor diodes are used for electronic tuning. The Karoo Array Telescope (KAT-7) requires a front-end lter that has the ability for ne-tuning the response after fabrication, in order to obtain the designed ltering response. Not only are tune-all characteristics required, but also wide bandwidth, at passband, high selectivity and implementation in microstrip technology. In this thesis an extensive investigation of both tunable and wideband lters is done, in order to nd a possible solution for the KAT-7 speci cations. Following this investigation, it is concluded that no suitable design approach for tune-all wideband lters, implemented in microstrip, exists in current literature. Therefore, this thesis proposes a new type of lter along with the development of a complete design procedure. Two lters are designed with this procedure to achieve the required passband from 1.2 to 1.95 GHz (i.e. a fractional bandwidth of 49%). In the rst lter design, with a network consisting of 4 cascaded ltering-sections, the centre frequency is 5% tunable and the bandwidth 17.5%. With the second lter consisting of 6 cascaded ltering-sections, higher selectivity is achieved but with lower return loss. Here the centre frequency is 8.5% tunable and the bandwidth 18.8%. The theoretical results are validated with the fabrication of both lters. This design is very unique in that it achieves wide bandwidth, is realisable in microstrip and most importantly is tunable in both the centre frequency and bandwidth. An advantage of this design procedure is that full wave simulations are minimal, due to the complete circuit models used for optimisation. / AFRIKAANSE OPSOMMING: 'n Nuwe soort van wye-band lters, met verstelbaarheid in beide senter frekwensie en bandwydte, word voorgelê in hierdie tesis. Hierdie lters is gebaseer op versteurde ringresoneerders in kaskade, terwyl varaktordiodes gebruik word vir elektroniese verstelling. Die Karoo Array Telescope (KAT-7) vereis 'n voorkant lter wat die vermoë het vir die instemming van die respons na fabrikasie, sodat die geontwerpde lter respons behaal kan word. Nie net word verstel-als eienskappe vereis nie, maar ook wye bandwydte, plat deurlaatband, hoë selektiwiteit en implimentering in mikrostrook tegnologie. In hierdie tesis is 'n veelomvattende ondersoek gedoen van beide verstelbare en wyeband lters, sodat 'n moontlike oplossing vir die KAT-7 spesi kasies gevind kan word. Na aanleiding van hierdie ondersoek, is die gevolgtrekking dat daar geen gepaste ontwerp benadering vir verstel-als wye-band lters, wat geïmplimenteer is in mikrostrook, in huidige literatuur bestaan nie. Daarom stel hierdie tesis, saam met die ontwikkeling van 'n volledige ontwerp prosedure, 'n nuwe tipe lter voor. Twee lters is ontwerp met hierdie prosedure om die vereiste deurlaatband vanaf 1.2 tot 1.95 GHz (dit is 'n fraksionele bandwydte van 49%) te behaal. In die eerste lter ontwerp, met 'n netwerk wat uit 4 kaskade lter-seksies bestaan, is die senter frekwensie 5% verstelbaar en die bandwydte 17.5%. Met die tweede lter bestaande uit 6 kaskade lter-seksies, word hoër selektiwiteit behaal maar met laer eggoverswakking. Hier is die senter frekwensie 8.5% verstelbaar en die bandwydte 18.8%. Die teoretiese resultate is geldig bewys deur die fabrikasie van albei lters. Hierdie ontwerp is baie uniek in dat dit wye bandwydte behaal, is realiseerbaar in mikrostrook en mees belangrikste dat dit verstelbaar is in beide senter frekwensie en bandwydte. 'n Voordeel van hierdie prosedure is dat heelgolf simulasies minimaal is, a.g.v. die volledige stroombaan modelle wat gebruik word vir optimering.

The problem with which this investigation was concerned was that of determining the use of modern automotive tune-up equipment as used by the automotive garages in the Fort Worth area. The data for this study were provided by fifteen garages of the Fort Worth area. Of the Fort Worth garages using electronic equipment to perform a tune-up, few appear to have the adequate equipment and few appear to make more use of the equipment. In addition, a deficiency seems evident in the use of the electronic computer in performing tune-ups.

In der vorliegenden Arbeit wurde mit einer Reihe von Experimenten geprüft, ob sich die Tonarterinnerung von Nichtabsoluthörern durch aus der Gedächtnisforschung abgeleitete Einflussfaktoren erklären lässt. Zunächst erfolgte eine theoretische Betrachtung des Tonartgedächtnisses sowohl aus musikpsychologischer als auch aus gedächtnispsychologischer Perspektive. Die Analyse von Befunden zum „latenten“ und „echten“ absoluten Gehör zeigte, dass eine Reihe von potenziellen Einflussfaktoren auf die Tonarterinnerung betrachtet werden muss, um herauszufinden, ob es sich bei diesen beiden Phänomenen um unterschiedliche Ausprägungen derselben Fähigkeit handelt. Um den Einfluss von Faktoren der Melodien, der Melodie-Lernenden und der Art des Melodie-Lernens auf die Tonarterinnerung zu prüfen, wurden insgesamt 268 Probanden gebeten, vertraute Melodien aus dem Gedächtnis zu singen. Unabhängige Variablen waren die musikalische Expertise der Probanden, ihre Fähigkeit Töne zu benennen, die Form und die Intensität des dem Experiment vorangegangenen Melodie-Lernens sowie verschiedene Charakteristika der Melodien. Abhängige Variable war die Genauigkeit, mit der die Originaltonarten der Melodien produziert wurden. Es konnten Effekte der Hör-Häufigkeit, der musikalischen Expertise, der Tonbenennung, der Melodie-Eingängigkeit sowie ein Effekt motorischer Kontextinformationen auf die Genauigkeit der Tonarterinnerung nachgewiesen werden. Um den Häufigkeitseffekt mit einer weiteren Anforderung zu untersuchen, wurde in einem weiteren Experiment die Tonbenennungsleistung von Absoluthörern und Nichtabsoluthörern verglichen. Dabei kam die Methode der Pupillometrie zum Einsatz, um Unterschiede in der mentalen Beanspruchung beim Benennen von Tönen unterschiedlicher Klangfarbe und Tonklasse nachweisen zu können. Die Ergebnisse stützen die Annahme, dass das häufige Hören bestimmter Töne sowohl bei Absoluthörern als auch bei Nichtabsoluthörern die Tonbenennung erleichtert. Dies verweist darauf, dass auch bei der musikspezifischen Aufgabe der Tonbenennung ein so grundlegendes Prinzip des menschlichen Gedächtnisses wie die Stabilisierung von Gedächtnisinhalten durch Wiederholung zum Tragen kommt. Insgesamt weisen die Ergebnisse darauf hin, dass Tonarterinnerung ein komplexes Phänomen ist, für das eine alleinige Erklärung als „latentes absolutes Gehör“ zu kurz greift. Statt einer schwachen Ausprägung einer hochspezialisierten Fähigkeit scheint es sich eher um eine eigene Form des Erinnerns, die auf allgemeingültigen Gedächtnisprinzipien beruht, zu handeln. / In this thesis, memory for musical keys among absolute pitch nonpossessors, which is often referred to as “latent” absolute pitch, is examined. A theoretical analysis focused on existing research about “latent” and “manifest” absolute pitch. Evidence from music-psychological and general memory research as well as neuropsychological evidence was considered. The review of existing research revealed that several factors are potentially relevant for the memory of musical keys and should be considered in trying to determine whether “latent” and “manifest” absolute pitch can be described as different levels of the same ability on an “absolute pitch continuum”. To examine whether characteristics of learned melodies, of melody-learners, and of melody-learning influence memory for musical keys among absolute pitch nonpossessors, 268 participants were asked in a series of experiments to sing familiar melodies from memory. Independent variables were the musical expertise of participants, their ability to label pitches, type and intensity of melody-learning, and characteristics of the learned melodies. The accuracy with which learned melodies could be produced in the original key was the dependent variable. Results revealed that frequency of melody-learning as well as participants’ musical expertise and ability to label pitches influence the accuracy of key production. Whether or not a melody is catchy as well as the existence of different types of motor imagery are further influencing factors for the accuracy of key production. To examine the frequency-of-hearing effect in more detail, another experiment compared the pitch labeling performance of absolute pitch possessors and nonpossessors. Pupillary responses were measured in order to show differences in mental resource allocation when labeling pitches of different key colors or timbres. Results support the assumption that frequent exposure to pitches of certain key colors or timbres facilitate their labeling among both absolute pitch possessors and nonpossessors. This suggests that basic principles of human memory such as learning by frequency of exposure affect also very specific tasks such as pitch labeling. Taken together, the results suggest that memory for musical keys is a complex phenomenon which can not adequately be described as being simply a “latent” or weak form of absolute pitch. Instead, memory for musical keys can be described as a “normal” memory mechanism, influenced by factors known to influence numerous other forms of human memory.

Submitted by Sandra Pereira (srpereira@ufrrj.br) on 2016-07-28T13:42:46Z No. of bitstreams: 1 2011 - Luciano Martins.pdf: 2085328 bytes, checksum: e82ea56c9ce66ac46b97ad748fb1de5a (MD5) / Made available in DSpace on 2016-07-28T13:42:46Z (GMT). No. of bitstreams: 1 2011 - Luciano Martins.pdf: 2085328 bytes, checksum: e82ea56c9ce66ac46b97ad748fb1de5a (MD5) Previous issue date: 2011-02-25 / Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior - CAPES / This study aims to compare three sampling techniques to estimate species richness of ants (Hymenoptera: Formicidae) in ?campos of altitude? at Itatiaia National Park, in Rio de Janeiro state. Compared richness estimates for genera and species, and curves of accumulation of species for sardine baits, pitfall traps and manual collecting are presented for two distinct sites ("steppe vegetation" [VE] and "rupicola vegetation" [VR]) in Itatiaia Plateau. The three techniques collected 31 species of 10 genera belonging to four subfamilies. In VE were registered 50% of genera and 36.4% of species collected exclusively by baits, and 20% and 50% by pitfall traps, respectively. In VR were registered 37.5% of genera and 47.61% of species exclusive for baits, but no one genus and species was restricted for manual collecting. In the case of ?campos de altitude? and when only one technique can be used, the sardine bait was the most efficient sample technique to registered richness at VE. For VR, the manual collecting was the most appropriate technique to record species richness. However, the pitfalls (in VE) and bait (in VR) recorded an important portion of species richness. In our data, the manual collecting were associated with standardized data collecting protocols, and, although much criticized, have shown effectiveness in recording species in habitats where widespread techniques cannot be used. In spite of different physiognomies show differentiated efficiency for samples techniques, it is relevant to using more than one technique associated to a colleting protocol well structured to potentiated the record of the new species in environments as the ?campos de altitude?. / O presente trabalho tem por objetivo comparar tr?s t?cnicas de coleta para estimar a riqueza de esp?cies de formigas (Hymenoptera: Formicidae) no Planalto do Itatiaia , Rio de Janeiro, Brasil. S?o apresentados dados comparativos das estimativas da riqueza de g?neros, de esp?cies e as curvas de acumula??o de esp?cies amostradas com iscas de sardinha, armadilhas de queda e coletas manuais, em duas fisionomias distintas (?vegeta??o est?pica? [VE] e ?vegeta??o rup?cola? [VR]), no Planalto do Itatiaia. Com a utiliza??o das tr?s t?cnicas foram coletadas 30 esp?cies de 11 g?neros, pertencentes a quatro subfam?lias. Para a VE foram registrados 50% dos g?neros e 36,4% das esp?cies coletadas exclusivamente pelas iscas, e 20% e 50% pela armadilha de queda, respectivamente. Para a VR foram registrados 37,5% e 47,61% dos g?neros e esp?cies exclusivas para iscas, por?m nenhum g?nero e esp?cie foram restritos a coleta manual. Em se tratando de ecossistema de campos de altitude e quando apenas uma das t?cnicas possa ser utilizada, a isca de sardinha foi a mais eficiente para amostrar a riqueza em VE. Para VR, a coleta manual se apresentou como a t?cnica mais apropriada para registrar sua riqueza. No entanto, as armadilhas de queda (na VE) e as iscas (na VR) registraram uma parcela n?o t?o desprez?vel da riqueza. Em nossos dados, as coletas manuais foram associadas a protocolos de coleta padronizados, e, apesar de serem bastante criticadas, mostraram efici?ncia no registro de esp?cies em habitats em que outras t?cnicas mais difundidas n?o podem ser utilizadas. Apesar de fisionomias distintas apresentarem efici?ncia de t?cnica de coleta diferenciada, ? relevante a utiliza??o das tr?s t?cnicas para se potencializar ainda mais o registro das esp?cies em ambientes de campos de altitude.

El fin de este proyecto de tesis fue Sintonizar un controlador PID, por los métodos de Ganancia Límite y Tanteo, para controlar de manera remota la variable Nivel en un módulo educativo del Laboratorio de control de la Universidad Ricardo Palma, esto se hizo con la finalidad que el alumno pueda entender la sintonía de un PID, lo cual es siempre recurrente en el campo. El control remoto permitió asemejar mucho a los procesos actuales. Al finalizar la tesis se obtuvo un control PID sintonizado bajo el método de Ganancia Límite con valores de Kp=0.6, Ti=0.067 minutos y Td=0.017 minutos; el cual dio valores de Error de estado estacionario=3.41% o 0.341, Tiempo de estabilización=30 segundos, Máximo sobre impulso=5.505% y Tiempo de subida=26 segundos; para un Set Point de 10 cm. Por otro lado, para un Set Point de 30 cm, dio valores de Error de estado estacionario=3.03% o 0.91, Tiempo de estabilización=67 segundos, Máximo sobre impulso=0.381% y Tiempo de subida=66 segundos. Además, se sintonizó bajo el método de Tanteo con valores de Kp=0.3, Ti=0.019 minutos y Td=0.01 minutos; el cual dio valores de Error de estado estacionario=2.99% o 0.299, Tiempo de estabilización=40 segundos, Máximo sobre impulso=6.69% y Tiempo de subida=39 segundos; para un Set Point de 10 cm. Por otro lado, para un Set Point de 30 cm, dio valores de Error de estado estacionario=1.87% o 0.56, Tiempo de estabilización=75 segundos, Máximo sobre impulso=0.398% y Tiempo de subida=73 segundos. The purpose of this thesis project was Tuning a PID controller, using methods of grope and Gain Limit to remotely control the variable level in an educational module Control Laboratory of the University Ricardo Palma, this was done in order that students can understand the tuning of a PID, which is always recurring in the field. The remote control allowed much resembles the current processes. After the thesis tuned PID control method gain under the limit values of Kp = 0.6 was obtained, Ti = Td = 0067 minutes and 0017 minutes; which gave values of steady state error = 3.41% or 0.341, Settling time = 30 seconds Maximum overshoot = 5.505% and Rise Time = 26 seconds; Set for a Point of 10 cm. On the other hand, for a set point of 30 cm, gave values of steady state error = 3.03% or 0.91 Settling time = 67 seconds Maximum overshoot = 0.381% and Rise Time = 66 seconds. Furthermore, it was tuned under trial and error with values of Kp = 0.3, Ti = Td = 0019 minutes and 0.01 minutes; which gave values of steady state error = 2.99% or 0.299, Settling time = 40 seconds Maximum = 6.69% overshoot and rise time = 39 seconds; Set for a Point of 10 cm. On the other hand, for a set point of 30 cm, gave values of steady state error = 1.87% or 0.56 Settling time = 75 seconds Maximum overshoot = 0.398% and Rise Time = 73 seconds.

This research presents an innovative approach to extracting information from Positive Energy Districts (PEDs), urban areas generating surplus energy. PEDs are integral to the European Commission's SET Plan, tackling housing challenges arising from population growth. The study refines BERT to categorize PED-related entities, producing a cutting-edge NER model and an integrated pipeline of diverse NER tools and data sources. The model achieves an accuracy of 0.81 and an F1 Score of 0.55 with notably high confidence scores through pipeline evaluations, confirming its practical applicability. While the F1 score falls short of expectations, this pioneering exploration in PED information extraction sets the stage for future refinements and studies, promising enhanced methodologies and impactful outcomes in this dynamic field. This research advances NER processes for Positive Energy Districts, supporting their development and implementation.