172
Arşivlenen Tez
0
DOI Atanmış
0%
DOI Oranı
Arşivlenen Tez
Semantic relation extraction by enriching word embeddings exploiting turkish morphology
Distributed representations (DR) are used to capture semantic and syntactic patterns in language by analyzing the distributional relationships of words within textual data. The modeling methods that produce DR are based on the assumption (distributional hypothesis) that "words that occur in the same context tend to have similar meanings," which is inherent to the nature of language. These modeling methods, due to their unsupervised nature, can be trained without human judgment input, allowing researchers to train large datasets at relatively low costs. Although word-based models perform effectively for languages with limited vocabularies, such as English, they exhibit considerable inefficiency when applied to morphologically rich languages with unlimited vocabularies, such as Turkish. We observed that n-gram and statistical segmentation methods, which are commonly used in subword modeling to address the issues of out-of-vocabulary and rare-words, are highly sensitive to orthographic similarity. Consequently, these methods struggle to distinguish between unrelated concepts (e.g., shrink - shrine). Moreover, we noted that the impact of morphological segmentation methods on these types of problems has shown inconsistent results in the literature. This thesis aims to make conceptual assumptions and improvements concerning different types of semantic relationships (e.g., relatedness and similarity), to model the role of language morphology as an input in subword DR models, and to develop the dataset generation methodologies and evaluation methods to measure this effect. Within the scope of the study, different models and segmentation methods were empirically tested, the AnlamVer and OSimUnr datasets were produced, and the task of relatedness classification and associated evaluation methods were proposed to measure the noise introduced by segmentation to the model. Our experiments demonstrate that morphological segmentation produces significantly less noise compared to n-gram-based methods and can lead to substantial performance improvements depending on the nature of the task.
Comparative study for earthquake performance of steel buildings with seismic isolator and fixed based steel buildings with damper
Earthquake causes huge loss of lives and enormous damages to properties every year. In order to understand and avoid such damages, different types of seismic isolators have been used. To get the optimal and effective types of seismic isolators, a comparison study of three same-sized 12-storey steel buildings with conventional steel braced frames, and lead rubber bearing (LRB) fixed-based with fluid viscous dampers (FVD) is conducted, and their seismic performance enhancements are evaluated using SAP2000 software. Researchers suggest that the seismic isolator building can survive seismic agitation behaviors such as uplifts, stress against ruptures, shears, cracking and displacements. In the study, three same-sized steel structures with different seismic isolator models, special moment and concentric moment frames of all three structures (SMF and SCBF) respectively, earthquake location (Kocaeli Turkey, Yarimca, 8/17/1999) have been selected, performing non-linear time history analysis, non-linear evaluation of dynamic behavioral building response spectrum analysis under load varying time function and design parameters are conducted. As a result of this study, conventional buildings, lead rubber bearing buildings and fluid viscous damper building storey displacement, inter-storey drifts, mode shape, shear force, axial force, base shear, and time history analysis for nonlinear dynamic structural responses are evaluated and compared with the conventional, lead rubber bearing and fluid viscous damper building, according to American Institute of Steel Construction (AISC 360-16). Keywords; Lead Rubber Bearing, Conventional Structure, Fluid Viscous Damper, Fixed Base, Response Spectrum, and Nonlinear Time History Analysis; SAP2000
Multiband and concurrent matching network design
Multiband network design has become a big issue due to the current improvements in RF and MW technologies. Since the new wireless protocols come up fast such as 3G, 4G, 5G etc., the RF and MW hardware must be designed to provide services to all distinct technologies in simultaneous regimes. To avoid interference, noise, harmonics, etc., the use of multiband matching or filtering at the front-end or back-end of RF/MW devices is an inevitable work. Particularly multiband matching in a multiband system is always necessary to be sure of proper power transfer between each multiband units. In this thesis, the multiband matching network design problem is addressed under the analytical perspective. Therefore, a deep investigation on network theoretical foundations is performed. In addition, only passive networks are concerned for providing concurrent operation. One of the well-known semi-analytical network design methods, which is called real frequency technique, is taken as a main domain to develop the thesis contributions. Two main approaches are elaborated. The first approach is the generation of proper frequency mapping functions and determination of prototype networks to design multiband matching network. The proposed frequency transformation approach is useful and practical for multiband filter and matching network designs. As a second research track of the thesis; a synthesis based approach is aimed to be developed for generalized multiband matching network designs. A proper integration of the proposed approach with RFT matching techniques and examination of the implementation results are provided.
Procurement and activity planningfor a production line provider
Consider a production line provider with the following problem. Turnkey production lines should be delivered to global clients by contractual due-dates. Machines in the production lines can be bought from suppliers or built in-house. Since the production lines for different clients are of the same type -albeit of different capacity- they consist of similar or same machines. Thus, their purchase and/or production need to be coordinated to reduce costs and production time. Furthermore, capital and workforce resources are constrained. Hence, not all activities can be done in parallel and should be spread across the planning horizon. Delivery delays have monetary fines proportional to the time past the due-dates. The company is concerned about making the due-dates by observing all resource constraints and carefully coordinating the activities of the concurrent projects. We give mixed-integer linear programming formulations for the investigated problems and report the results of the numerical experiments conducted using these formulations.
KeNet: A comprehensive Turkish wordnet and its applications in text clustering
In this thesis, we summarize the methodology and the results of our efforts to construct a comprehensive WordNet for Turkish. Most languages have access to comprehensive language resources. Traditional resources like bilingual dictionaries, monolingual dictionaries, thesauri, and lexicons are developed by lexicographers. As computer processing of languages gains popularity, a new set of resources become necessary. One such resource is WordNet which was initially constructed for the English language at Princeton University. A WordNet contains much of the information contained in a classic dictionary, but it also contains additional relationship information. These relations go beyond synonym relation and give information about relations such as a word being"is-a" or "is-a-part-of" another. These semantic relations are used in many text analysis tasks. A WordNet also categorizes words under common concepts. These concepts are called as synsets. As a result of all these, WordNet is a comprehensive dictionary which is readable by the computers and a useful language resource for text analysis and other research based on human language. In Turkish language, our WordNet is not the first. The previous WordNet is part of BalkaNet project which is a multilingual WordNet including Turkish and Balkan languages. BalkaNet contains only common words between these languages, as such BalkaNet does not contain all Turkish words and suffers from top-down constructing method disadvantages. BalkaNet project has not been updated or expanded in recent years. In this work, we construct a Turkish WordNet from scratch using a bottom-up method. In general, there are two methods for constructing WordNets. Bottom-up method means that we create the WordNet from scratch while top-down approach uses other WordNets by translating them. We use Turkish Contemporary Dictionary (CDT) which is an online Turkish dictionary provided by Turkish Language Institute. Bottom-up approach has its own difficulties since constructing a WordNet from scratch requires more resources and a lot of effort. In this work, we extract synonyms from CDT and ask experts to match common meanings for pairs of synonyms. We developed an application which makes annotation step easier and more accurate. We also use two groups of annotators to measure inter-annotator agreement. We used some automatic approaches to extract semantic relations from Turkish Wikipedia (Vikipedi) and Vikisözlük. We processed CDT to extract candidate synonyms and used rule-based approaches to find synonym sets. There is no thesaurus for Turkish, so as an application we construct a thesaurus automatically and measured accuracy with our manually constructed synsets. We named our WordNet "KeNet". Finally, in this thesis, we developed a novel approach to represent a text document in a vector space. This approach uses WordNet semantic relations. This part of the thesis is an application of KeNet. We used our approach to represent text documents and implemented two different clustering algorithms over these vectors. We tested our method over Turkish Wikipedia articles, domains of which are labeled by Wikipedia.
An inverse source problem connected thermoacoustic imaging in multi-layer planar medium
In this thesis, we mentioned imaging technics used today, and microwave induced thermoacoustic procedure and imaging. We made literature survey about solution of the thermoacoustic equation for homogeneous and inhomogeneous medium. We modelled the inhomogenity of medium as a multi-layer planar structure and de- ned initial condition, continuity conditions on the layer boundaries and radiation conditions at in nity, then we derived analytical forward and inverse solution of the thermoacoustic wave equation for inhomogeneous medium with the source distribution existing in all layers. Our solution of inverse source problem is based on the methods of the Green's functions for layered planar media. For qualitative testing and comparison of the point-spread functions associated with the conventional solution for homogeneous medium and our derived layered solutions, we performed numerical simulations. In numerical simulations rst we generated the measured data by using the derived forward solution for multi-layer planar medium and then we used conventional inverse solution and our derived inverse solution to image source distribution. Our simulation results showed that the conventional inverse solution based on homogeneous medium assumption, as expected, produced incorrect locations of point sources, whereas our inverse solution involving the multi-layer planar medium produced point sources at the correct source locations. Also, we showed that the performance of layered inverse solution is sensitive to the validity of the layer parameters and medium parameters used as prior information in the measured data. Our inverse solutions based on multi-layer planar media are applicable for cross-sectional 2 dimensional imaging of the organs such as breast, skin, and abdominal structure.
Analyzing the benefits and challenges of Building Information Modelling and Life Cycle Assessment integration
The architecture, engineering and construction (AEC) industry involves the large quantities of material and energy consumption. Previous studies show that the AEC industry consumes about 40% of primary energy utilization and contributes up to more than half of global GHG emissions. With the aim of decreasing these negative impacts of AEC industry on the natural environment, the integrated use of advanced technological instruments has been increasing in the last decade. One of the promising technological instrument integrations is the collaborative use Building Information Modeling (BIM) and Life Cycle Assessment (LCA) which is able to decrease negative impacts of AEC industry on natural environment in an effective and efficient manner. The objective of this MS thesis is to identify, classify and prioritize the benefits and challenges of the integrated use of Building Information Modelling and Life Cycle Assessment in the AEC industry. In order to achieve research objectives of this thesis, a comprehensive literature review, semi-structured interviews with experts and Delphi method were performed. The interrater agreement and significance-level statistics were conducted for analyzing and validating the consensus among the Delphi method participants. Results of this study provide twenty-two types of benefits and seven types of challenges for the integrated use of BIM and LCA in the AEC industry. The contributions of this thesis are a comprehensive identification, classification and prioritization of the integrated use of BIM and LCA. The results of this study may contribute to extend utilization of BIM-LCA integration in the AEC industry.
Metamutator: Its realizations and its applications
Mutators became very popular after 1971, when Leon Chua realized the memristor, his postulated fourth circuit element, using them. The reason for the popularity of mutators, which are easily realizable 2-port devices, lies in the fact that they render possible the simulation/emulation of, hence experimentation with, postulated non-existing (not off the shelf available) elements like memstors . On the other hand, in the literature when simulating/emulating elements with 2-ports, many 4-port "generalized mutator-like" realizations that nobody has been able to identify are being used. These underlying 4-ports, and their ability to act as a mutator when two of the ports are properly terminated, have been thus named metamutator. In this thesis, in addition to introducing some of metamutator realizations, newly designed metamutator circuits with one or two active devices are introduced. Also, a new active device, with only twelve transistors, named Additive and Differential IC (AD-IC), is proposed with its layout and use in metamutator circuit design. In addition, many 1-port and 2-port circuit realizations using metamutators have been introduced. 1-port applications are memstor simulation/emulation, floating and/or grounded impedance scaling which comprise inductance simulation, capacitance multiplication, oscillators and Frequency Dependent Negative Resistor (FDNR) simulation. 2- port applications cover Voltage Mode Multiple Input Single Output (VM-MISO) and Current Mode Single Input Multiple Output (CM-SIMO) universal filters and implementations of transconductance and transimpedance amplifiers. Also, two different applications of AD-IC: AD-IC based analog multiplier and AD-IC based full-wave rectifier are proposed in the thesis.
Ring array processing for forward-looking intravascular and intracardiac ultrasonic imaging
Forward-looking (FL) catheter-based imaging systems are highly desirable for guiding interventions in IVUS and ICE applications. One of the main challenges in the array based FL-IVUS systems is the large channel count which results in increased system complexity. Synthetic phased array processing with reduced firing count simplifies the front-end and hence can enable 3-D real-time imaging. Recently, we have investigated dual-ring arrays suitable for IVUS and ICE imaging. In this dissertation, we present and explored two different optimized array designs based on dual circular and hexagonal rings that are suitable for synthetic phased array processing with reduced number of firings. We used simulated annealing to optimize the wideband PSF of the array system. The optimal firing set from different Tx-Rx elements eliminates most of the redundant spatial frequencies in the coarray and produces low side lobes and number of firings. The optimization procedure starts with the calculated wideband PSF of the full set coarray including all Tx-Rx firing combinations. In every iteration step, a predefined coarray with only one element change is compared with the previous set in terms of cost function. For the two designs, we obtained optimal reduced set by minimizing both the peak side lobe level and the integrated side lobe energy. For two test cases, we obtained reduced sets of 512 and 256 coarray elements, and constructed 2-D PSFs. The simulation results show that the optimized dual circular and hexagonal ring arrays provide up to 10-dB and 5-dB improvements, respectively, in peak near side lobe level with no widening in main lobe width. Our test results show that the side lobe levels of the dual circular and hexagonal arrays can be remarkably and controllably reduced by using simulated annealing optimization, and the resulting optimal array designs can meet the resolution and speed requirements of 3-D real time IVUS and ICE imaging.
Solid-state arrays and beamformers for side-looking intravascular ultrasonic imaging
Current solid-state intravascular ultrasound (IVUS) devices use a one-dimensional(1-D) rectangular array wrapped around the cylindrical catheter. Beamforming inthese devices is limited to synthetic phased array processing, and the image quality isdegraded by low signal to noise ratio (SNR) and lack of focusing in elevation. Wepropose new solid-state reconfigurable cylindrical array configurations, based onnon-uniform aperture sampling, enabling realization of low-cost integrated phasedarray beamformers based on the latest IC and transducer integration techniques. Weemployed non-uniform (Fresnel) aperture sampling with identical differential delaysbetween adjacent elements to reduce dynamic beamformer complexity. For thispurpose, the angles of the element boundaries and centers are determined byequalizing path difference of neighboring elements. For Side Looking-IVUS (SLIVUS),the image plane is scanned through stepping the active aperture, and theactive array must be reformed in each step. Fresnel sampled ring array leads to asimplified phased array beamformer with N/2 identical differential delays. The samelogic applies to a two-dimensional (2-D) array with NL×NE-elements over acylindrical patch. We compared the proposed and existing array designs throughnumerical point spread function (PSF) simulations. The uniform and Fresnel sampled1-D ring arrays perform similarly in terms of beam quality while the latter improvesimage SNR by a factor of ?NL due to synthetic vs full phased array processing. TheFresnel sampled 2-D array performs dramatically better in elevation compared to 1-Dring array. The additional contribution of 2-D array on image quality is high SNRproduced by phased array beamforming. The proposed Fresnel sampling ofcylindrical aperture simplifies phased array beamformer by halving the requireddelay element count and transforms the complex 2-D beamformer into a separablebeamformer that requires simple 1-D beamformers. The resulting 1-D and 2-D SLIVUSarray configurations could be realizable using monolithic integration of theanalog beamformers with transducer arrays.