Dissertations / Theses on the topic 'Appr. automatico'

Le fotocamere digitali acquisiscono, manipolano e salvano in memoria informazioni elettronicamente tramite un sensore e unitá di calcolo integrate, rendendo la fotografia accessibile agli utenti finali i quali non necessitano di basi di conoscenza di chimica o delle specifiche operazioni meccaniche che erano necessarie per lo sviluppo con la fotografia analogica. Diversi tipi di degradazioni e artefatti possono affliggere le immagini acquisite tramite camere digitali, riducendo la fedeltà percepita delle immagini e rendendo piú difficili operazioni di visione artificiale che possono essere operate sulle immagini acquisite. Tre elementi possono essere identificati come possibili sorgenti di artefatti in un'immagine: il contenuto della scena, le limitazioni hardware della camera e la pipeline di elaborazione dell’immagine, dall’acquisizione al salvataggio in memoria. Alcuni artefatti non sono direttamente trattati all’interno della tipica pipeline di camera, come ad esempio la presenza di nebbia o pioggia che possono ridurre la visibilitá della scena acquisita. Questi artefatti , per essere trattati, necessitano di metodologie ad hoc che sono generalmente applicate come operazioni post-processing. Altri tipi di artefatti sono legati ai processi di elaborazione dell’immagine e all'intera pipeline di camera integrata nelle camere. Questi artefatti includono il rumore proveniente dal sensore, dominanti cromatiche indesiderate, basso contrasto e artefatti di compressione. L’obiettivo di questa tesi è l'identificazione e la progettazione di nuovi e piú robusti moduli di elaborazione e restauro dell’immagine capaci di migliorare la qualitá delle immagini acquisite dalle camere digitali, in particolare in scenari critici quali condizioni climatiche avverse, condizione di bassa illuminazione etc… Gli artefatti identificati sono divisi in due macro gruppi: “artefatti in-camera” e “artefatti esterni” Nel primo gruppo sono stati identificati e trattati quattro argomenti: la rimozione del rumore proveniente dal sensore della camera, il processo di bilanciamento automatico del bianco, il miglioramento automatico del contrasto e la rimozione di artefatti di compressione JPEG. Il processo di progettazione di questi moduli ha tenuto conto di aspetti di efficienza, in termini di calcolo e memoria occupata, in relazione alla possibilità di integrare tali moduli in future configurazioni di pipeline di camera. Il secondo gruppo di artefatti è legato alla presenza di elementi nella scena che possono causare una degradazione in termini di fedeltà dell’immagine e/o usabilità. In particolare il lavoro presentato nella tesi è focalizzato su artefatti provenienti dalla presenza di pioggia nella scena scattata. La tesi, dopo una descrizione della tipica pipeline di elaborazione delle camere digitali, analizza i diversi tipi di artefatti che possono affliggere la qualità di una immagine, e descrive la progettazione delle soluzioni proposte. Tutti gli approcci proposti sono basati su tecniche di apprendimento automatico, come ad esempio Reti Neurali Convoluzionali o procedure di ottimizzazione Bayesiana, e sono stati validati sperimentalmente su dataset standard. I contributi principali di questa tesi possono essere riassunti in tre punti: integrazione di approcci classici di imaging con tecniche di ottimizzazione basate su machine learning, progettazione e sviluppo di nuovi approcci e architetture di deep learning per il restauro e l'elaborazione di immagini e analisi di metodi di image processing basati su deep learning in task di computer vision.
Digital cameras record, manipulate, and store information electronically through sensors and built-in computers, which makes photography more available to final users which do not anymore need to rely on the use of chemicals and knowledge of mechanical procedures to develop their pictures. Different types of degradation and artifacts can affect images acquired using digital cameras, decreasing the perceptual fidelity of images and making harder many image processing and analysis tasks that can be performed on the collected images. Three elements can be identified as possible sources of artifacts in an image: the scene content, the hardware limitations and flaws, and finally the operations performed by the digital camera processing pipeline itself, from acquisition to compression and storing. Some artifacts are not directly treated in the typical camera processing pipeline, such as the presence of haze or rain that can reduce visibility of the scene in the depicted images. These artifacts require the design of ad hoc methods that are usually applied as post-processing on the acquired images. Other types of artifacts are related to the imaging process and to the image processing pipeline implemented on board of digital cameras. These include sensor noise, undesirable color cast, poor contrast and compression artifacts. The objective of this thesis is the identification and design of new and more robust modules for image processing and restoration that can improve the quality of the acquired images, in particular in critical scenarios such as adverse weather conditions, poor light in the scene etc… . The artifacts identified are divided into two main groups: “in camera-generated artifacts" and “external artifacts and problems". In the first group it has been identified and addressed four main issues: sensor camera noise removal, automatic white balancing, automatic contrast enhancement and compression artifacts removal. The design process of the proposed solutions has considered efficiency aspects, due to the possibility of directly integrating them in future camera pipelines. The second group of artifacts are related to the presence of elements in the scene which may cause a degradation in terms of visual fidelity and/or usability of the images. In particular the focus is on artifacts induced by the presence of rain in the scene. The thesis, after a brief review of the digital camera processing pipeline, analyzes the different types of artifacts that can affect image quality, and describes the design of the proposed solutions. All the proposed approaches are based on machine learning techniques, such as Convolutional Neural Networks and Bayesian optimization procedure, and are experimentally validated on standard images datasets. The overall contributions of this thesis can be summarized in three points: integration of classical imaging approaches with machine learning optimization techniques, design of novel deep learning architectures and approaches and analysis and application of deep learning image processing algorithms in other computer vision tasks.

La tesi tratta l'utilizzo del machine learning per la classificazione e la predizione di attacchi ad un software. In particolare sono stati sviluppati: l'applicazione presa di mira, ovvero un sito di cloud storage; un programma Python per rilevare gli attacchi e uno per automatizzarli; uno script per il training e l'utilizzo di un classificatore SVC.

La tesi si occupa di creare un modello di classificazione analizzando dataset riguardanti le Human Activity Recognition. Il lavoro di tesi è stato diviso in due parti. La prima parte tratta lo stato dell'arte, in cui sono stati analizzati e approfonditi i temi di Machine Learning e sistemi HAR. Nella seconda parte è stato descritto il progetto. Il progetto si basa sul processo di Data Mining. Durante tutto il processo di Data Mining sono state fatte valutazioni e sperimentazioni sia durante la fase di pulizia dei dati e di preprocessing, sia durante la fase di classificazione valutando vari algoritmi. Alla fine del processo di Data Mining è stato ottenuto un modello di classificazione soddisfacente. Infine, è stata sviluppata un' applicazione moblie in Android, che accedendo ai valori di due sensori (accelerometro e giroscopio) e servendosi del modello di classificazione già creato, è in grado di riconoscere l'attività svolta dall'utente.

Generating C code from MATLAB can be a powerful tool for a wide range of engineering problems. Especially in cases where MATLAB excels, including matrix operations and some of the built-in functions, code conversion may represent a significant assistance in their development. The aim of this master's thesis is to explore the possibilities of generating C code using MATLAB Coder App. As the documentation for MATLAB Coder was not found sufficient a theoretical basis for the use of this tool was introduced along with simple examples of possible applications. Next, code generation for controlling a laboratory DC motor using a microcontroller dsPIC was made automatic and controlled directly from a single MATLAB script. Individual steps were described to supplement existing documentation and to allow for an easy use. Finally, selected functions were tested for code generation. As a result, certain limitations can be outlined, as well as suitability of both the created tool and the MATLAB Coder App.

Despite being explored and constantly improved through the years, onboarding of new hires in corporate organizations has remained a challenge. Many of the issues can be linked to a lack of communication between the organization and the new employee, as well as the common nature of these environments where information is spread across job titles and sources. This thesis discusses the feasibility of implementing a basic chatbot that will allow new hires to ask questions and request varied information at all times, using an interface such as a messaging app. This research explores the way chatbots should be designed in order to be effective, reliable and enjoyable from a user experience perspective. The chatbot was developed using the Chatfuel platform and tested by new employees at a corporate environment. The users were requested to explore the chatbot freely and then asked to answer a survey. The interactions were also recorded and analyzed from in both qualitative and quantitative ways (chat logs and analytics). The study proves that an onboarding chatbot is a useful tool for new hires and can be used as a communication facilitator between the organization and the new hires during the first weeks of employment, and also after that, serving as an information source and a broadcasting method. The chatbot gives the new hires an accessible source of information that helps on the process of getting to speed, and enables a positive experience that increases familiarity in the new workplace.

碩士
開南大學
資訊管理學系
103
App development for smart phones and tablets is one of the hottest technologies in recent years. As the app can be downloaded and updated via the Internet, the revision of the software is much faster than the traditional PC software. How to use tools to improve development efficiency and program quality becomes more and more important. Testing is a necessary phase in software development. This paper discusses how to use JUnit Framework to test Android apps automatically, including the creation of testing classes and test cases. We use three different types of Apps in our experiments. One is the Contact App that uses the SQLite database access functions to save the profile data; the second is a Service, whichcan execute in the background; the third app is a Content Provider app. We implemented automatic test classes and test cases for them, and compared the efficiency of automatic testing and human testing. Experimental results show that automatic testing can significantly reduce the time spent on testing and reduce the physical and mental burden of testers. Besides, automatic testing can also avoid erroneous operations. The test cases can also be used in the future when the program is modified, and they can be used as auxiliary program documentation. Our study shows that the automatic testing technology can effectively improve the development efficiency and the app quality.

Every day, developers add new applications (apps) to the Google Play Store, which ease users' lives and entertain them. The rapid development of these apps is only possible through the provision of software libraries whose functionality can be directly integrated into an app without creating it from scratch. For instance, some libraries provide extended possibilities for displaying content or additional support for specific networking capabilities. All these libraries are bundled with the main application code into one binary to avoid delays due to the loading of external functionality. The wide distribution of Android apps and the high turnover in this market attracts criminal actors (attackers). These attackers decompile the apps, integrate additional ad libraries or malware, and republish them. Through this approach, they use the apps’ popularity to trick users into downloading their repackaged apps. To prevent such malicious practices, developers and producers of libraries have begun to obfuscate their apps to make the decompilation process more challenging. However, the obfuscation of apps is not only done by developers but also by attackers to make the detection of copyright infringement harder or hide their malicious intent. While analysts try to protect the developers’ copyright and the privacy and security of app users, code obfuscation hinders them from identifying libraries and repackaged apps, detecting obfuscated names and strings, and recovering them. The obfuscated code might contain not only malware but also vulnerabilities or unauthorized access to private data. This dissertation introduces different approaches that support the analyses mentioned above using static analysis, dynamic analysis, and machine learning. Since the obfuscation of repackaged apps makes it difficult to distinguish between library and app code, we present approaches for library detection, separation of app code, and mapping of library code. We evaluated the effectiveness of these approaches under the influence of different obfuscation techniques. Furthermore, we present our approach for identifying repackaged apps that uses our library identification to measure the similarity between repackaged and original apps without the influence of library code, which would distort the measurement. Further contributions support the recovery of names from obfuscated entities in library code. Finally, we presented an approach for identifying and recovering obfuscated strings that supports data-flow analyses. Using our approaches, we outperformed all state-of-the-art competitors. Furthermore, we analyzed in total over 100,000 apps for obfuscated names, obfuscated libraries, obfuscated strings, and repackaged apps.

碩士
國立彰化師範大學
資訊工程學系
102
The influence of smartphones and Apps on our life habits and consumption behavior is heavy nowadays. In addition to business websites, Apps are required for commercial firms to provide services and promote products. The construction of Apps often suffers some problems： the cost of developing and maintaining Apps is high, the operating systems of smartphones are various, and the possibility that information is inconsistent on websites and Apps. To alleviate the mentioned problems, this paper uses component-based software engineering to automate the construction of websites and Apps. First, component-based software engineering is used to establish software components for business function and commercial firms can use the activity diagram to compose function components. An automated software development approach is proposed to reuse existing software components for constructing business websites and Apps. The cost of software development can thus be reduced. Second, the web pages of HTML5 are utilized to be the user interface for composing and displaying function components. The HTML5 language can be executed on both PCs and smartphones, and most operating systems and browsers support HTML5. Once some data in the database server is updated, information acquired from websites and Apps is consistent and up to date. Finally, the generated websites and Apps are deployed in the cloud system. Commercial firms maintain data directly in the cloud system. There is no need for commercial firms to buy and maintain servers, databases, and software. Therefore, the cost of maintenance is reduced.

碩士
國立臺灣大學
電機工程學研究所
102
Software development of mobile computing and cloud services presents a paradigm shift. In this new paradigm, we may often need to deal with system under test (SUT) without source code and up-to-date documentation of the SUT. And when it comes to software testing in the industry, engineers often spend lots of effort to produce test cases themselves, writing script or manipulating SUT. Here we introduce a framework of automatic test case generation procedure that creating SUT trace, building the model of an SUT out of execution traces, and generating test cases according to the built model with specifications in Linear-time Temporal Logic (LTL). Then executing the test cases to check the SUT and examining the validation of test cases. As long as all transitions in a fail trace have been observed in the test session, then our procedure will eventually generate a test case that corresponds to the fail trace. We implement and report experiments on Android mobile applications.

碩士
國立臺灣大學
企業管理碩士專班
107
Depression is a common but serious mental illness that affects the ability to think, and daily activities causing patients to suffer every day, and even committing suicide. From the statistics at World Health Organization, depression affected 350 million people worldwide, and WHO listed depression as the fourth most significant cause of suffering and disability worldwide [1]. Over 800,000 suicide every year, making it the second leading cause of death in 15-29 years old. In Taiwan alone, 2 million people had suffered from depression, and 1.27 million people now are taking long-term antidepressants by statistics of Taiwan Ministry of Health and Welfare [2]. Depression accounts for the biggest share of the world’s burden of disease, measured by years lost to disability (YLD): healthy years ‘lost’ because they are lived with a physical or mental disability. When ranked by disability and death combined, depression comes ninth behind prolific killers such as heart disease, stroke and HIV. Depression is caused by combination of genetics, biological, environment and psychological factors, and is treatable even for the most severe depression. Yet depression is widely undiagnosed and untreated because of social stigma, unawareness of the disease and lack of mental health resources [3]. To further explore the problems of insufficient of treatment of depression, we conduct interviews with depressed patients and therapists in Taiwan. From the interviews, we conclude that the reasons behind are high cost of mental treatment, low treatment availability, patients unaware of their depression severity level, and social stigma. Lack of assessment and awareness of depression result in insufficient and ineffective care for the disease despite the effective psychological and pharmacological treatments for depression. With the advance of 5G, the smartphones combined with cloud computing break through the limited computational power in smartphones. The raise of awareness of mental health in millennials results in prevalence of metal health APPs. We therefore propose a solution to diagnose depression with the help of facial recognition, speech processing, and natural language processing utilizing artificial intelligence. With the interface of mobile phone APP, our solution could easily access to everyone. This goal is to detect depression in the early stage to prevent further loss in financial and personal life, and urge our customers to go to professional clinics or hospitals for further help if diagnosed as severe or moderate depression. Automatic detection of depressive symptoms would potentially increase the rate of diagnostic visit, and patients’ awareness of their own depression severity levels, leading to faster intervention. In the research paper conducted by Professor Fei-Fei Li at Stanford University, automatic AI based depression severity test could achieve both accuracy and efficiency utilizing artificial intelligence [4]. The automatic detection algorithms identify facial traits and voices characteristics could help provide a universal and low-cost way of spotting the early signs of depression with smartphones. In practice, clinicians identify depression in patients by first measuring the severity of depressive symptoms during in-person clinical interviews. During these interviews, clinicians assess both verbal and non-verbal indicators of depressive symptoms including monotone pitch, reduced articulation rate, lower speaking volumes, fewer gestures, and more downward gazes [4]. If such symptoms persist for two weeks, the patient is considered to have a major or severe depressive episode. Structured questionnaires such as DSM-IV [5] and PHQ-9 [6] have been developed and validated in clinical populations to assess the severity of depressive symptoms. The major purpose of our APP is to serve potential patients with convenient diagnosis to increase their self-awareness of their own depression severity levels in an accessible, affordable, and effective way. Also, the therapists and psychiatrists could use the App to monitor their patients and it serves as marketing channels for related fields such as hospitals, clinics, pharmaceutical companies, medical device companies, gaming, recreational, entertainment, and exercising App companies.

碩士
國立臺灣大學
電子工程學研究所
105
Automatic software testing of Android applications becomes more and more important. Fast and extensive testing activities have always been the targets for testing providers. In this thesis, we develop a software tool called ADAS to manage devices in a device farm and schedule testing tasks to increase the throughput, efficiency, and reliability of testing activities. Our experiment shows that our tool can bring benefits to software testing activities in many aspects.

碩士
崑山科技大學
數位生活科技研究所
101
The study proposes to use an Android-based smart phone as home automation monitor, and creates on the home control server on the monitor a user-friendly interface, where users will be able to control home appliances through the user-friendly interface. The monitor would transmit control signal to Bluetooth hardware gateway for the conversion of RF wireless control signal through Bluetooth device to control lamp, door lock, electric fan, air-condition and other home appliance devices through a manner of RF wireless transmission. The study even combined temperature and door lock access sensor applications. Users would be able to set up the range of air-condition and electric fan automated temperature control and the monitor would activate electric fan and air-condition as per temperature number detected. The Monitor would ring the doorbell to notify users of visitors coming once visitors press the doorbell, while access control would able to detect whether gangsters would like to break into the house maliciously, and then the monitor will send a text message and notify user through Gmail proactively. In order to allow users to understand the status of residence anytime and anywhere, the study uses i-jetty micro web server and installed it in the monitor, thus smart phone can be used to connect to the monitor from other places through internet, to monitor situation and control home appliance devices at home anytime.

碩士
南開科技大學
車輛與機電產業研究所
103
In many countries, there are more and more users of smart handheld device (phones, and tablets), the rapid development of APP program has become the software that those users commonly used. This study aimed to develop the APP program with smart handheld device, we use the online tool of APP inventor to develop the interface ordering system. In the traditional restaurant, to order a meal, it is still by a waiter (waitress) to help customer finish his or her order. This research is to change the traditional paper order into the APP program order. When the customer holds the smart handheld device to the restaurant, there is a QR Code (Quick Response Code) on the restaurant's door or window can quickly download the restaurant's APP program, and customers use this APP programs can clearly know how the menu and inside situation with restaurant. Also, while the customers come into the restaurant after waiter directs him or her to the table, customers can just use the APP program of smart handheld device to see restaurant menu, and the customer finish to order his or her meals by the APP program , it will send information by bluetooth device to the counter and kitchen, the cashier just click the APP program to let them see the customer's order, and understand the amount and the turn of order to accelerate the time of make a meal and reduce errors during process. Also, the table number and the meals of table are displayed in the counter, too. It can avoid miss information of meals. The APP program of smart handheld device offers a lot of convenience and comfortable for customers, also can increase the willingness to come again. The smart handheld device can be equipped with more and more devices and also combine with other media, to become an important way to attract customer.

ZigBee home automation systems have been around for years. They include smart energy efficient wireless systems that connect devices via mesh networking. It has always been a problem to connect to these ZigBee nodes with existing technologies we use daily. The premise of this project is to provide easy access to these nodes via near field communication, which is a technology that is available on a majority of smart phones today. The report details the design and implementation of N2Z, a device that translates data provided by an Android app through a NFC interface to a ZigBee device and vice- versa. We first discuss the requirements and use cases for the device and application. Then the report goes on to discuss the base technologies behind N2Z, which include the Android platform, Near Field Communication (NFC) & the ZigBee platform. The report then goes on to discuss the design and implementation of the system. Consequently, we discuss the final results of the system with certain use cases and provide some empirical data that gives us insight into system operation. Next we discuss other existing technologies that exist in the market. Finally we conclude the report with a discussion on future work and thoughts on the future of ZigBee.
text

碩士
國立臺灣大學
電機工程學研究所
106
Nowadays, increasing number of developers are using software testing technology to ensure the quality of their products in both web and mobile applications. It is a technique that providing random event traces and observing to see if the system or applications crash or not. At the moment, analyzing testing traces is still a heavy time-consuming and labor-intensive work. When processing software testing, we follow two testing protocols: one is to randomly execute actions on the page and the other is to set up modules of actions. For the result, the former one we can know that whether the service or system is able to overcome this kind of stress test and the later one is to test if our defined functions work well. However, both behaviors require setting targets and labeled them ahead. For human beings, it is such an easy job to classify or detect contents but for software testing, it is not. And this is the reason why we still require labor force for doing so. In this thesis, we present Smart-Eye, a visualized analysis service for both web pages and mobile applications. With the advantage of image classification, Smart-Eye helps software testers in organizing data and labeling them for the later testing scripts. Not only the analysis but the analysis improves the time consumption and accuracy. Because there is no complete dataset in this domain, we first gathered needed information and then built a classifying model from them. Last, we used it to do our intense work and made it an auto-labeling tool for later software testing in order to speed up the process of software testing in both aspects above.

Internet keeps growing everyday and with that, the creation of new web pages. Due to this fact, web pages of many different categories can be found such as News, Sports or Business. This issue has made investigators think about one innovative concept: Webpage Classification. This new approach implies the categorization of web pages to one or more category labels. Some research has been done during the last years using text and visual content extracted from the web pages to be able to classify. However, the need of being able to do such a thing in an Android app has not been investigated yet, to the best of our knowledge. Consequently, this thesis is focused in the development of an Android app which is able to classify web pages. First of all, text and visual features have to be extracted from each webpage. Four types of visual features were extracted from each web page to construct a visual features vector of 160 attributes. Concerning to the text features, a text features vector was also built for each of the webpage with 160 attributes. To do so, a “Bag-Of-Words” of one hundred and sixty words was set up from the HTML code already extracted and filtered. Thus, we end up having a full vector of 320 attributes for each webpage. A binary classification was performed trying to distinguish web pages for Adults and for Kids. Good results were obtained especially when using AdaBoost classifier with text and visual features where a 94.44% of accuracy of correct classifications was achieved.

碩士
國立臺灣師範大學
工業教育學系
104
The purpose of this study was to investigate the relationships between engineering APPs and learning effectiveness amongst college students. The sample included 233 students from Taipei and New Taipei taking a course on automatic control. A questionnaire was developed to collect data which was then subject to statistical analysis. Major conclusions of the study were summarized in the following. In the course on automatic control: 1. Student’s learning motivation was above average. 2. Most students adopted pragmatist learning style. 3. Student’s learning motivation was middle level. 4. Student’s learning motivation, App usability, and learning outcome varied due to different background variables. 5. Student’s learning motivation was significantly correlated to learning style, App usability, and learning outcome. Summarizing the findings listed above, this study provided corresponding recommendations to course instructors, colleges, students in engineering majors, APP developers, as well as future researchers.

The Internet of Things has accelerated the availability of solutions for smart homes, o ering an increasingly varied range of products and smart home applications that aim to simplify daily tasks and increase user comfort. However, the available systems are often excessively complex from the user's perspective, making their use unintuitive and unpredictable. Within the scope of this dissertation, an application for smart homes was developed based on the Flutter framework, which can interact with Altice labs' Smart Home device ecosystem. This document presents a comparison between the various existing smart home solutions, as well as an overview of the solution developed by Altice labs. The implementation of the proposed solution is detailed using code excerpts that illustrate the choices made in terms of design and functionality. Finally, the developed application is evaluated using a test process that veri es compliance with its requirements.
A Internet das Coisas acelerou a disponibilização de soluções para casas inteligentes, oferecendo uma gama cada vez mais variada de produtos e de aplicações de domótica inteligente que visam simplificar as tarefas do dia-a-dia e aumentar o conforto dos utilizadores. Contudo, frequentemente os sistemas disponibilizados são excessivamente complexos na perspectiva do utilizador, tornando a sua utilização pouco intuitiva e imprevisível. No âmbito desta dissertação foi desenvolvida uma aplicação para casas inteligentes baseada na framework Flutter que é capaz de interagir com o ecossistema de dispositivos Smart Home da Altice labs. Este documento apresenta uma comparação entre as diversas soluções de casas inteligentes existentes, além da visão geral da solução desenvolvida pela Altice labs. A implementação da solução proposta é detalhada recorrendo a excertos de código que ilustram as opções tomadas em termos de design e de funcionalidade. Por fim, a aplicação desenvolvida é avaliada recorrendo a um processo de teste que verifica o cumprimento dos seus requisitos.
Mestrado em Engenharia Eletrónica e Telecomunicações

The range of applications for high-quality automatic speech recognition (ASR) systems has grown dramatically with the advent of smart phones, in which speech recognition can greatly enhance the user experience. Currently, the languages with extensive ASR support on these devices are languages that have thousands of hours of transcribed speech corpora already collected. Developing a speech system for such a language is made simpler because extensive resources already exist. However for languages that are not as prominent, the process is more difficult. Many obstacles such as reliability and cost have hampered progress in this regard, and various separate tools for every stage of the development process have been developed to overcome these difficulties. Developing a system that is able to combine these identified partial solutions, involves customising existing tools and developing new ones to interface the overall end-to-end process. This work documents the integration of several tools to enable the end-to-end development of an Automatic Speech Recognition system in a typical under-resourced language. Google App Engine is employed as the core environment for data verification, storage and distribution, and used in conjunction with existing tools for gathering text data and for speech data recording. We analyse the data acquired by each of the tools and develop an ASR system in Shona, an important under-resourced language of Southern Africa. Although unexpected logistical problems complicated the process, we were able to collect a useable Shona speech corpus, and develop the first Automatic Speech Recognition system in that language.
MIng (Computer and Electronic Engineering), North-West University, Potchefstroom Campus, 2014

Internship Report presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics
Nowadays, there is a greater urgency for companies to innovate and digitally transform. Automation and digitalization of processes, coming up with new ways to connect with employees and customers, or investing in a robust infrastructure that can respond and extract insights from the ever-increasing amount of data that is being produced across all sectors of the enterprise: these are all topics that any company needs to have as a top priority if they want to keep up with their competitors and stay relevant in their own markets. The present document describes the activities carried out during a period of 10 months (September 2019 – June 2020, correspondent to the fiscal year of 2020) at Microsoft’s Corporation Portuguese Subsidiary, working as a full-time Technical Specialist. The technical specialist first received extensive technical and commercial training from a vast variety of resources (online resources, one-to-one shadowing, and on-site technical readiness) and after the ramp-up process, he proceeded to drive the business for Microsoft Portugal by conducting customer and partner meetings, and to play his part in the sales motion by providing deep technical expertise and compelling technology demonstrations. The technical specialist oversaw a unified set of Microsoft technologies, called the Power Platform, that bundle together a set of tools such as Power BI for data analysis and visualization, Power Apps for line-of-business application development, Power Automate for workflow automation and RPA capabilities and Power Virtual Agents as an engine for creating and deploying intelligent chatbots. All these components are further enriched and integrated with the Office 365 and Dynamics 365 ecosystems, hundreds of data connectors, advanced database capabilities, artificial intelligence and machine learning models, and the extensibility from the resources available on the Azure Cloud.