Sharif Digital Repository

Keyphrase extraction, as an important open problem of Natural Language Processing (NLP), is useful as a stand-alone task in the field of Information Extraction and as an upstream task for Information Retrieval, text summarization and classification,etc. In this study, regarding the needs in Persian NLP, artificial neural networks are adopted to extract keyphrases from single documents and a graph-based re-scoring method is proposed for multidocument keyphrase extraction. The proposed method for extracting keyphrases from multiple documents consists of two steps: (1) extracting keyphrases of each document in a cluster using a sequence to sequence model with attention, and (2) re-scoring the... 

This thesis examines the modal characteristics of the payload after connecting to the carrier via the adapter. Basically, the modal characteristics of a payload change after connection to the carrier due to the different stiffness of the coupling mechanism. In this report, it is attempted to split and model the assembly structural system into two substructures interconnected by the V band clamp mechanism. Then, each of the substructures analyze individually and the modal characteristics of the integrated system are extracted using the infrastructure coupling methods. The coupling method is the "Component Mode synthesis" (CMS) method and input data, dynamic and modal characteristics required... 

In recent years, word embeddings as the word representation have captured the attention of natural language processing (NLP) researches. One of the great advantages of word embeddings is their capability in representing the relationships of the words. Therefore, using word embeddings in NLP applications results in better performance.Despite widespread attention towards word embedding in late years, Persian word embeddings have not achieved sensible progress. One of the Persian word embeddings difficulties is related to that, Persian is a low-resource language in comparison with worldwide languages. Therefore, Persian word embedding quality is lower than English. Consequently, the accuracy of... 

This project has been conducted in two related phases: In the first phase, we have attempted to write a program capable of answering thematic similarity multiple-choice questions without utilizing any training data. The best performance in this phase was attained by the 25-topic LDA model using the Hellinger distance between the probability distributions of the poetic verses. This model managed to attain an accuracy of 42%, which is very close to the average human performance of 43%. In the second phase, two tasks of seven-class classification and binary classification were defined based on the p-value of the questions. To this end, the questions were initially ranked according to the... 

News plays a very important role in stock market trading. Nowadays news from a different part of the world and about different fields can be accessed easily, and for a successful trade, it is necessary to analyze accurately and use this big data and information as soon as possible. For this reason, this thesis tries to present and study models based on Deep Learning networks and Natural Language Processing for financial news analysis and predicting stock indices movement. This research takes advantage of a language model for learning and representing news text, and beside this language model it uses deep learning networks at multiple levels to extract proper features from each news in a day... 

In this thesis, the mechanical properties of collagen hydrogels were characterized using Finite Element method and the collagen content effect on the mechanical properties of hydrogels were investigated. Hydrogels samples with different collagen content were assessed which caused diverse mechanical behaviors. Due to the nonlinear behavior of hydrogels, using numerical methods and simulation softwars can be helpful to determine the mechanical properties of these material and save agreat deal of time. Finite Viscoelaticity theory was exploited for a UMAT Subroutine in Abaqus and an special Starin Energy Function was selected to extract the formulations. The agreement between simulation results... 

The purpose of this thesis is sliding mode control of a roll-pitch seeker by considering the zenith pass problem. This problem takes place when the target appears in front of the seeker’s head. First, the mathematical model of the roll and pitch frames is obtained by the Newton-Euler equations of motion. Then, a MIMO sliding mode controller is designed. The performance of the controller is investigated in the presence of model uncertainties and seeker disturbances. The results indicate that the designed controller is robust enough and provides a suitable performance in the stabilization, tracking and guidance loops.
 

The flexible magnetic robot is used for minimally invasive surgeries where there is a complex environment. Precise control of the position of the end of the robot and quick adaptation of the robot to uncertainties are among the most important challenges in this field, where the controller is responsible for compensating the error and controlling the position of the end of the robot. Now, is it possible to create a kind of memory in the system that when faced with the same or similar errors in the same situations or close to previous errors, the system uses its past and compensates the error? The importance of this work is that the response speed of the system is increased and the system can... 

Multi-disciplinary design optimization (MDO) of a surface to air flying vehicle has been done in this research, on the basis of flight simulation. The MDO problem have five disciplines consists of aerodynamic, propulsion, guidance, control, and fire control. Multiple campaign scenarios are defined. The MDO problem has 31 design variables and two objective functions. The objective functions are flying vehicle weight and miss distance. A new meta-heuristic algorithm has been designed to solve multi-objective optimization problems. The new algorithm has been called Multi-objective Adaptive Real-coded Memetic Algorithm (MARCOMA). The MARCOMA can solve large-scale and multi-objective problems and... 

PVDF (polyvinylidene fluoride) magnetic nanocomposite films were prepared by solution casting using two types of nanomagnetic particles: (i) magnetic nanoparticles synthesized by using an alkaline solution of mixed ferrous/ferric salts along with oleic acid as a coating agent at 70˚C; (ii) Fe3O4 nanoparticles prepared by hydrolysis of aqueous solution of ferrous/ferric salts with potassium hydroxide at room temprature. Scanning electron microscopy (SEM) showed that the PVDF magnetic nanocomposite had a porous structure with Fe3O4 nanoparticles dispersed inside the porous polymer matrix. The presence of crystalline magnetite within the polymer matrix was confirmed by X-Ray diffraction method... 

A cooperative aerial system to defense a Ground Station (GS), against an Incoming aerial Targets (IT) is presented. GS is surrounded by given terrains and a group of homogenous Unmanned Aerial Vehicles (UAVs) are employed using a novel online guidance algorithm in a decentralized manner. The proposed algorithm includes loiter, midcourse and terminal phases. During loiter; UAVs follow an optimal circular path. IT is supposed to approach GS along an optimal low altitude trajectory with respect to the terrains. UAVs are informed the initial position and velocity of IT and they are unaware of IT trajectory. Each UAV decides on whether to engage with IT or not, and shares its decision with other... 

In this research, multi-disciplinary design optimization (MDO) of a flying vehicle has been done based on the flight simulation. A meta-heuristic algorithm called Multi-objective Adaptive Real-coded Memetic Algorithm (MARCOMA) has been used for optimization. Since solving a MDO problem is a time consuming process, a RBF neural network has been used in the optimization algorithm as a surrogate model. The new algorithm, called MARCOMA+NN, has been tested with some standard benchmarks. MDO problem has six disciplines consists structure, aerodynamic, propulsion, guidance, control, and fire control. The MDO problem has 31 design variables and two objective functions. The objective functions are... 

Topic that is investigated in this article include identify and localization a number of aerial target penetrating into a protected by area by a swarm of UAVs is fixed wing. UAVs do not have any prior notice of positioning targets. In addition, it is assumed that lacks any radiation exploitable targets and thus use UAVs for Targets Detection from an active radar system. Obviously this article focus on developing a guidance algorithm UAVs to this particular issue and also as the targets participatory is localization algorithm. In addition, extraction type and performance characteristics radar system required and also performance characteristics of the class of UAVs, which is suitable to... 

In this project, a method for cooperative search and localization of RF ground moving targets by a group of UAVs is developed. It should be noted that UAVs are just equipped with GPS and directional sensors. Since there is fuel constraint for UAVs, they take fuel from a tanker whenever they require. Moreover, searching method enables the UAVs to see different parts of the desired area with almost uniform distribution. In addition, the proposed approach enables the UAVs to perform a local search with the aim of finding the targets having lost their signal during localization mode. Finally, based upon a fueling decision function the UAVs take turn, approach to the fuel tanker, and start... 

Nowdays much of the organic solid waste and wastewater is buried using centralized and traditional methods. This leads to many problems, including groundwater pollution, methane emissions into the atmosphere and rapid transmission of disease. In order to avoid these problems, using a decentralized approach can be a good solution. In this study, recovery of organic waste and wastewater of residential buildings using anaerobic digester as a decentralized way, have been examined. Using experiments, the optimum mass ratio of organic solid waste and wastewater is obtained as 80% and 20% respectively. Biogas production model is developed using mass balance and Balance of electrons in redox... 

The purpose of this paper is to design an integrated controller and observer (ICO) for a nonlinear system using genetic programming. ICO is a function that constructs control command directly from the measured state variables of the system. It means that, this function should imitate the behavior of the observer and controller and control the system with acceptable performance in different initial conditions, at the presence of disturbances and system uncertainties. The complexity of this design method, is not related to the complexity of the plant, in fact, the complexity in plant just effects at run time, but the design procedure does not change. So, if exact model of plant exist, using... 

In this research a novel intelligent back-stepping control method is developed. This method is robust to sensor noise and external disturbances. In addition, the controller is robust under model uncertainty. This controller does not need precise knowledge of system parameters. This method is based on three methods of: back-stepping control, least squares estimation and a fuzzy compensator. This controller is used to control quadrotor stand witch is like an inverse pendulum. In quadrotor stand modeling, the inverse pendulum effect is considered too, witch is one of the innovations of the research. By doing various simulations, the validity of controller is tested. Also the performance of the... 

In this study, a multiple model wind estimator is proposed to detect the wind type and to estimate the wind components as well as the states of a fixed-wing UAV without any direct measurement of the air data. Then, the identified wind model and the estimated states are compensated in the heuristic nonlinear model predictive controller during landing phase. For this purpose, a static multiple model approach is taken, comprised of four independent extended Kalman filters, each one is estimating the wind based on one of the four wind models including a constant wind, a “1-cosine” model, a wind shear and a microburst.Moreover, a new heuristic multiple model filter, called Multiple Model Extended...