Speaker-based language identification for Ethio-Semitic languages using CRNN and hybrid features.

Natural language is frequently employed for information exchange between humans and computers in modern digital environments. Natural Language Processing (NLP) is a basic requirement for technological advancement in the field of speech recognition. For additional NLP activities like speech-to-text translation, speech-to-speech translation, speaker recognition, and speech information retrieval, language identification (LID) is a prerequisite. In this paper, we developed a Language Identification (LID) model for Ethio-Semitic languages. We used a hybrid approach (a convolutional recurrent neural network (CRNN)), in addition to a mixed (Mel frequency cepstral coefficient (MFCC) and mel-spectrogram) approach, to build our LID model. The study focused on four Ethio-Semitic languages: Amharic, Ge'ez, Guragigna, and Tigrinya. By using data augmentation for the selected languages, we were able to expand our original dataset of 8 h of audio data to 24 h and 40 min. The proposed selected features, when evaluated, achieved an average performance accuracy of 98.1%, 98.6%, and 99.9% for testing, validation, and training, respectively. The results show that the CRNN model with (Mel-Spectrogram + MFCC) combination feature achieved the best results when compared to other existing models.

Software defined networking based network traffic classification using machine learning techniques.

The classification of network traffic has become increasingly crucial due to the rapid growth in the number of internet users. Conventional approaches, such as identifying traffic based on port numbers and payload inspection are becoming ineffective due to the dynamic and encrypted nature of modern network traffic. A number of researchers have implemented Software Defined Networking (SDN) based traffic classification using Machine Learning (ML) and Deep Learning (DL) models. However, the studies had various limitations such as encrypted traffic detection, payload inspection, poor detection accuracy, and challenges with testing models both in offline and real-time traffic modes. ML models together with SDN are adopted nowadays to enhance classification performance. In this paper, both supervised (Logistic Regression, Decision Tree, Random Forest, AdaBoost, and Support Vector Machine) and unsupervised (K-means clustering) ML models were used to classify Domain Name System (DNS), Telnet, Ping, and Voice traffic flows simulated using the Distributed Internet Traffic Generator (D-ITG) tool. The use of this tool effectively manages and classifies traffic types based on their application. The study discussed the dataset used, model selection, implementation of the model, and implementation techniques (such as pre-processing, feature extraction, ML algorithm, and model evaluation metrics). The proposed model in SDN was implemented in Mininet for designing the network architecture and generating network traffic. Anaconda Python environment was utilized for traffic classification using various ML techniques. Among the models tested, the Decision Tree supervised learning achieved the highest accuracy of 99.81%, outperforming other supervised and unsupervised learning algorithms. These results indicate that the integration of ML with SDN provides an efficient classification method for identifying and accurately classifying both offline and real-time network traffic, enhanced quality of service (QoS), detection of encrypted packets, deep packet inspection and management.

MPPT efficiency enhancement of a grid connected solar PV system using Finite Control set model predictive controller.

Maximum power point tracking (MPPT) is required to get the highest possible power generated from a photovoltaic (PV) cell. Numerous researchers have proposed different MPPT strategies to be able to collect maximum generated electricity from the photovoltaic cells. In this research paper, a MPPT model predictive control strategy for a grid-connected PV system is presented. Model predictive control (MPC) was used to develop and model the AC load energy tracking efficiency for the PV systems with a power rate of 20 kW at standard test conditions. For the purpose of obtaining the power tracking performance, a DC-DC boost converter, DC-AC two level three phase inverter, and control mechanism for a grid connected AC load system was examined and presented in this paper. To approximate the actual PV array properties, the PV model is used, and the MPPT approach is suggested as a way to regulate the DC-DC boost converter and get the most power possible from the PV array when compared to P&O and model predictive control system. A three-phase, two-level VSI is employed in this study that is controlled by a model predictive control system with SVPWM. The inverter's control structure is developed using a model predictive control system (inner loop current controller) with reference frame transformation (abc to dq) coordinates by utilizing PLL. The PLL is used to obtain critical information about the grid voltage. A RL filter is then used to lower the total harmonic distortion of the output and connect the inverter's output to the grid. The MATLAB R2019a environment is used to create the system model. The overall performance of the system for conventional perturb and observer is around 97.72%, while for Finite Control Set Model Predictive Controller is 99.80%, which is better than previous similar research with faster time response and less oscillation around maximum power point.

Derived Amharic alphabet sign language recognition using machine learning methods.

Hearing-impaired people use sign language as a means of communication with those with no hearing disability. It is therefore difficult to communicate with hearing impaired people without the expertise of a signer or knowledge of sign language. As a result, technologies that understands sign language are required to bridge the communication gap between those that have hearing impairments and those that dont. Ethiopian Amharic alphabets sign language (EAMASL) is different from other countries sign languages because Amharic Language is spoken in Ethiopia and has a number of complex alphabets. Presently in Ethiopia, just a few studies on AMASL have been conducted. Previous works, on the other hand, only worked on basic and a few derived Amharic alphabet signs. To solve this challenge, in this paper, we propose Machine Learning techniques such as Support Vector Machine (SVM) with Convolutional Neural Network (CNN), Histogram of Oriented Gradients (HOG), and their hybrid features to recognize the remaining derived Amharic alphabet signs. Because CNN is good for rotation and translation of signs, and HOG works well for low quality data under strong illumination variation and a small quantity of training data, the two have been combined for feature extraction. CNN (Softmax) was utilized as a classifier for normalized hybrid features in addition to SVM. SVM model using CNN, HOG, normalized, and non-normalized hybrid feature vectors achieved an accuracy of 89.02%, 95.42%, 97.40%, and 93.61% using 10-fold cross validation, respectively. With the normalized hybrid features, the other classifier, CNN (sofmax), produced a 93.55% accuracy.

Dynamics analysis of permanent magnet synchronous motor speed control with enhanced state feedback controller using a linear quadratic regulator.

This paper presents a robust speed regulation and control of a permanent magnet synchronous motor (PMSM). A linear quadratic regulator (LQR) based state feedback controller was developed to achieve a successful suppression of periodic disturbance of speed and torque. Sliding mode observer in conjunction with the disturbance observer was deployed in the control of motor speed. Simulations were carried out based on two compared controllers such as the state feedback controller and the conventional proportional-integral-derivative (PID) controller to attenuate the noisy effects of the external disturbance. A comparative analysis of results showed that a robust as well as an improved speed and torque dynamic performance was achieved with the state feedback (SFC) controller. A reduced periodic disturbance with percentage steady state error values of 24.17% and 23.51% was obtained with the SFC controller as compared to 38.0% and 38.37% obtained using a PID controller. The Eigen values obtained from the derived state feedback matrix (K) based on Ackerman's rule proved that the entire system operation is controllable and the performance index is marginally stable. All simulations were performed using MATLAB/SIMULINK version 2021.

DFT/TDDFT calculations of geometry optimization, electronic structure and spectral properties of clevudine and telbivudine for treatment of chronic hepatitis B.

Chronic hepatitis B remains a worldwide health concern. Presently, many drugs, such as Clevudine and Telbivudine, are recommended for the treatment of chronic hepatitis B disease. For this purpose, the quantum chemical analysis of ELUMO-HOMO (Egap), ionization potential (IP), electron affinity (EA), electronegativity (EN), chemical hardness (η), chemical potential (µ), chemical softness (S), electrophilicity index (ω), electron accepting capability (ω+), electron-donating capability (ω-), Nucleophilicity index (N), additional electronic charge (∆Nmax), Optical softness (σ0) and Dipole Moment, IR and UV-Vis spectra, molecular electrostatic potential (MEP) profile, Mulliken charge analysis, natural bond orbital (NBO) were examined in this study. The dipole moment of the compounds suggests their binding pose and predicted binding affinity. The electrophilic and nucleophilic regions were identified, and techniques such as NBO, UV-Vis, and IR were used to gain insights into the molecular structure, electronic transitions, and potential drug design for Hepatitis B treatment. Calculations for this study were carried out using the Gaussian 09 program package coupled with the DFT/TDDFT technique. The hybrid B3LYP functional method and the 6-311++G(d, p) basis set were used for the calculations.

Red pepper drying with a double pass solar air heater integrated with aluminium cans.

In this paper, an experimental evaluation of a newly developed flat plate double pass solar air heater combined with aluminum cans for drying red pepper was presented. The proposed solar dryer system was designed, modeled, and evaluated. Solar air heater trials were carried out using the absorber's top and bottom plate and aluminum cans for red pepper drying at Bahir Dar, Ethiopia. To test the solar dryer, 100 pieces of red paper were obtained from the Bahir Dar region of Ethiopia for the purpose of experimentation. Microsoft Excel was used to perform statistical analysis of eleven mathematical models. The results show that the mixed-mode solar greenhouse dryer takes less time to dry red pepper than the open solar dryer. In the midday, the solar insolation reached 973 W/m2 and the minimum solar insolation was 220 W/m2 and air is expelled at a rate of 0.0383 kg/s. According to the experimental results, the dryers chamber temperature ranged from 30.9 to 54 °C, while the ambient temperature was between 22.6 and 28.2 °C. The mixed-mode double pass achieves up to 46% and 28% efficiency when used with aluminum can dryers and conventional open sun dryers, respectively. A drying rate of 0.0003395 kg/s was achieved for the open sun dryer system and 0.0000365 kg/s for the mixed mode solar dryer. Using mixed-mode and open-sun solar dryers, the logarithmic model was found to be most effective in explaining the red pepper behavior. Furthermore, a comparison was made between the experimental and predicted moisture ratios through the calculation of the coefficient of determination (R2), the reduced chi-square (X2), and the root mean square error (RMSE). The results show that the logarithmic model achieved the highest value of the correlation coefficient (R2), which was determined to be 0.9978 and 0.9989, while the logarithmic model achieved the lowest value of Chi-square (X2).

Investigation of substituent effects on the electronic structure and antiviral activity of favipiravir derivatives for Covid-19 treatment using DFT and molecular docking.

In this study, Density-functional theory/Time-dependent density-functional theory (DFT/TDDFT) and Molecular docking method was used to investigate the effect of methyl acetate, tetrahydrofuran and cyanobenzylidene substituents on the electronic structure and antiviral activity of favipiravir for treating COVID-19. The DFT and TDDFT computations were employed using the Gaussian 09 software package. The values were calculated using the 6-311++G(d, p) basis set and the hybrid B3LYP functional method. Autodock vina software was used for simulations to better predictions and to validate the modified compounds' binding affinities and poses. Results of the study indicate that compounds 1 to 6 all displayed a planar structure, where the pyrazine ring, carboxamide, hydroxyl groups, and other substituents are all situated within the same plane. In addition, the energy gaps (Egap) of these six compounds (Cpd 1, 2, 3, 4, 5, and 6) were compared. The significant dipole moment and binding affinity achieved implies a particular orientation for binding within the target protein, signaling the anticipated strength of the binding interaction. In all six compounds, the electrophilic domain is situated in the vicinity of the amine functional group within the carboxamide compound, whereas the nucleophilic domain encompasses both the carbonyl and hydroxyl groups. The most negatively charged sites are susceptible to electrophilic interactions. In conclusion, compounds 5 and 6 exhibit a high binding affinity of the target protein, while compound 6 has a high energy gap, which could enhance its antiviral activity against the COVID-19 virus.

Geometry, reactivity descriptors, light harvesting efficiency, molecular radii, diffusion coefficient, and oxidation potential of RE(I)(CO)3Cl(TPA-2, 2'-bipyridine) in DSSC application: DFT/TDDFT study.

Dye-sensitized solar cells (DSSCs) are an excellent alternative solar cell technology that is cost-effective and environmentally friendly. The geometry, reactivity descriptors, light-harvesting efficiency, molecular radii, diffusion coefficient, and excited oxidation state potential of the proposed complex were investigated. The calculations in this study were performed using DFT/TDDFT method with B3LYP functional employed on the Gaussian 09 software package. The calculations were used the 6-311 + + G(d, p) basis set for the C, H, N, O, Cl atoms and the LANL2DZ basis set for the Re atom, with the B3LYP functional.. The balance of hole and electron in this complex has increased the efficiency and lifetime of DSSCs for photovoltaic cell applications. The investigated compound shows that the addition of the TPA substituent marginally changes the geometric structures of the 2, 2'-bipyridine ligand in the T1 state. As EDsubstituents were added to the compound, the energy gap widened and moved from ELUMO (- 2.904 eV) (substituted TPA) to ELUMO (- 3.122 eV) (unsubstituted). In the studying of solvent affects; when the polarity of the solvent decreases, red shifts appears in the lowest energy an absorption and emission band. Good light-harvesting efficiency, molecular radii, diffusion coefficient, excited state oxidation potential, emission quantum yield, and DSSC reorganization energy, the complex is well suited for use as an emitter in dye-sensitized solar cells. Among the investigated complexes mentioned in literature, the proposed complex was a suitable candidate for phosphorescent DSSC.

Control of a fixed wing unmanned aerial vehicle using a higher-order sliding mode controller and non-linear PID controller.

Unmanned aerial vehicles (UAVs) have seen a rise in use during the last few years. Such aircrafts are now a convenient way to complete dangerous, dirty, and tedious tasks. Given that their operation involves a control problem which is non-linear and coupled, it is difficult to analyse. This paper presents the modeling and control of a fixed-wing unmanned aircraft as a contribution to this field. The system's flight dynamics is derived using Newton's second law of motion. The system is designed to have a non-linear Proportional Integral Derivative (NPID) controller and a higher-order sliding mode controller (HOSMC). When simulating the system using MATLAB Simulink software, an external disturbance was added to test the robustness of the controllers. Five performance indices which include mean square error (MSE), integral time square error (ITSE), integral absolute error (IAE), integral time absolute error (ITAE), and integral square error (ISE), were used to compare the controllers performance. These indices are used to provide a numerical assessment of the two controllers' performance. The outcomes demonstrate that the roll, pitch, and yaw states performed better than the super-twisting sliding mode controller. On the airspeed control, the non-linear PID performed better than the super-twisting sliding mode controller.

Torque ripples enhancement of a PMBLDC motor propelled through quadratic DC-DC boost converter at varying load.

The characteristic effects of a permanent magnet brushless dc (PMBLDC) motor which is propelled by a quadratic DC-DC boost converter was investigated in this paper. The boost converter applied inductor coupling with semi-conductor switches which operate concurrently to produce a high conversion voltage ratio. Modeling of the PMBLDC motor was presented while simulation was performed using a closed loop control system with different controllers for the enhancement of torque-speed performance at varied load. Simulation results show that an optimum voltage gain with minimum voltage stress was achieved at 0.9 duty cycle while varying the inductor turns ratio (n) from 1 to 6 and inductor coupling coefficient (K) from 0.1 to 1.0 so as to ensure that the entire flux generated in one coil is fully linked to the other. A simplified PI controller was designed and experimented through simulation for excellent drive performance at varying load. The results from the spectral display show that the usual speed oscillation and torque ripples produced by the machine were attenuated using different speed controllers. The Total Harmonic Distortion (THD) values indicate that the PID controller achieved a minimal value of 24.06% for speed and 23.16% for torque as compared to 28.73% and 29.82% obtained from the PI controller, while the P controller achieved 41.88% and 35.67% as shown in the graphical abstract.

Novel predictive model of cell survival/death related effects of Extracellular Signal-Regulated kinase protein.

Computational modelling is a technique for modelling and solving real-world problems by utilising computing to provide solutions. This paper presents a novel predictive model of cell survival/death-related effects of Extracellular Signal-Regulated Kinase Protein. The computational model was designed using Neural Networks and fuzzy system. Three hundred ERK samples were examined using ten different concentrations of three input proteins: EGF, TNF, and insulin. Based on the different concentrations of input proteins and different samples of ERK protein, adjustment Anderson darling (AD) statistics for multiple distribution functions were computed considering different test such as visual test, Pearson correlation coefficient, and uniformity tests. The results reveal that utilising different concentrations and samples, values such as 7.55 AD and 18.4 AD were obtained using the Weibull distribution function for 0 ng/ml of TNF, 100 ng/ml of EGF, and 0 ng/mL of insulin concentrations. The model was validated by predicting the various ERK protein values that fall within the observed range. The proposed model agrees with the deterministic model, which was developed using difference equations.

Towards audio-based identification of Ethio-Semitic languages using recurrent neural network.

In recent times, there is an increasing interest in employing technology to process natural language with the aim of providing information that can benefit society. Language identification refers to the process of detecting which speech a speaker appears to be using. This paper presents an audio-based Ethio-semitic language identification system using Recurrent Neural Network. Identifying the features that can accurately differentiate between various languages is a difficult task because of the very high similarity between characters of each language. Recurrent Neural Network (RNN) was used in this paper in relation to the Mel-frequency cepstral coefficients (MFCCs) features to bring out the key features which helps provide good results. The primary goal of this research is to find the best model for the identification of Ethio-semitic languages such as Amharic, Geez, Guragigna, and Tigrigna. The models were tested using an 8-h collection of audio recording. Experiments were carried out using our unique dataset with an extended version of RNN, Long Short Term Memory (LSTM) and Bidirectional Long Short Term Memory (BLSTM), for 5 and 10 s, respectively. According to the results, Bidirectional Long Short Term Memory (BLSTM) with a 5 s delay outperformed Long Short Term Memory (LSTM). The BLSTM model achieved average results of 98.1, 92.9, and 89.9% for training, validation, and testing accuracy, respectively. As a result, we can infer that the best performing method for the selected Ethio-Semitic language dataset was the BLSTM algorithm with MFCCs feature running for 5 s.

Hybrid GWO-PSO based optimal placement and sizing of multiple PV-DG units for power loss reduction and voltage profile improvement.

Distributed generation (DG) is integrated in a passive distribution system to reduce power loss, improve voltage profile, and increase power output. To reap the most benefits of the distribution system, the best location and appropriate DG size must be determined. This paper presents a hybrid Grey wolf Optimizer (GWO) and Particle swarm optimization (PSO) approach for determining the best placement and DG size while considering a multi-objective function that includes active and reactive power loss minimization as well as voltage profile enhancement. Dilla distribution system was used as a case study and the weighted technique was used to convert to a single objective function while taking into account multiple constraints such as bus voltage limit, DG output limit, and branch current limit. DG penetration is limited to up 60% of the total active load on the feeder and a forward-backward sweep load flow algorithm was used to generate the load flow solutions. The findings of the study show that combining three PV-DGs (Case 3) is the best way to improve voltage profile and minimize losses. In addition, the proposed hybrid GWO-PSO algorithm performed better compared to the other four algorithms (Grey Wolf Optimization (GWO), Whale Optimization Algorithm (WOA), Particle swarm optimization (PSO), and sine cosine algorithm (SCA)) in terms of achieving the best multi-objective function (MOF) outcome.

Machine learning-based approach to predict ice meltdown in glaciers due to climate change and solutions.

The fate of humankind and all other life forms on earth is threatened by a foe, known as climate change. All parts of the world are affected directly or indirectly by this phenomenon. The rivers are drying up in some places and in other places, it is flooding. The global temperature is rising every year and the heat waves are taking many souls. The cloud of "extinction" is upon the majority of flora and fauna; even humans are prone to various fatal and life-shortening diseases from pollution. This is all caused by us. The so-called "development" by deforestation, releasing toxic chemicals into air and water, burning of fossil fuels in the name of industrialisation, and many others have made an irreversible cut in the heart of the environment. However, it is not too late; all of this could be healed back with the help of technology and our efforts together. As per the international climate reports, the average global temperature has increased by a little more than 1 °C since 1880s. The research is primarily focused on the use of machine learning and its algorithm to train a model that predicts the ice meltdown of a glacier, given the features using the Multivariate Linear Regression. The research strongly encourages the use of features by manipulating them to determine the feature with a major impact on the cause. The burning of coal and fossil fuels is the main source of pollution as per the study. The research focuses on the challenges to gather data that would be faced by the researchers and the requirement of the system for the development of the model. The study is aimed to spread awareness in society about the destruction we have caused and urges everyone to come forward and save the planet.

X-Ray image-based COVID-19 detection using deep learning.

COVID-19 is a type of respiratory infection that primarily affects the lungs. Obtaining a chest X-ray is one of the most important steps in detecting and treating COVID-19 occurrences. Our study's goal is to detect COVID-19 from chest X-ray images using a Convolutional Neural Network (CNN). This study presents an effective method for categorizing chest X-ray images as Normal or COVID-19 infected. We used CNN, activation functions dropout, batch normalization, and Keras parameters to build this model. The classification method was implemented using open source tools "Python" and "OpenCV," both of which are freely available. The acquired images are transmitted through a series of convolutional and max pooling layers activated with the Rectified Linear Unit (ReLU) activation function, and then fed into the neurons of the dense layers, and finally activated with the sigmoidal function. Thereafter, SVM was used for classification using the knowledge from the learning model to classify the images into a predefined class (COVID-19 or Normal). As the model learns, its accuracy improves while its loss decreases. The findings of the study indicate that all models produced promising results, with augmentation, image segmentation, and image cropping producing the most efficient results, with a training accuracy of 99.8% and a test accuracy of 99.1%. As a result, the findings show that deep features provided consistent and reliable features for COVID-19 detection. Therefore, the proposed method aids in faster diagnosis of COVID-19 and the screening of COVID-19 patients by radiologists.

CPD-CCNN: classification of pepper disease using a concatenation of convolutional neural network models.

Agricultural products are vital to the sustainability of the economies of developing countries. Most developing countries' economies such as Ethiopia heavily rely on agriculture. On a global scale, the pepper crop is one of the most important agricultural products in terms of human food security. However, it is susceptible to a variety of diseases which include blight leaf disease, gray leaf spot, common rust, fruit rot disease, powdery mildew symptoms on pepper leaf, and other related diseases that are all common today. Currently, more than 34 different pepper diseases have been discovered, resulting in a 33% average yield loss in pepper cultivation. Conventionally, farmers detect the disease using visual observation but this has its own demerits as it is usually not accurate and usually time consuming. In the past, a number of researchers have presented various methods for classifying pepper plant disease, especially using image processing and deep learning techniques. However, earlier studies have shown that binary classification requires improvement as some classes were more challenging to identify than others. In this study, we propose a concatenated neural network of the extracted features of VGG16 and AlexNet networks to develop a pepper disease classification model using fully connected layers. The development of the proposed concatenated CNN model includes steps such as dataset collection, image preprocessing, noise removal, segmentation, feature extraction, and classification. Finally, the proposed concatenated CNN model was evaluated, providing a training classification accuracy of 100%, validation accuracy of 97.29%, and testing accuracy of 95.82%. In general, it can be concluded from the findings of the study that the proposed concatenated model is suitable for identifying pepper leaf and fruit diseases from digital images of pepper.

Effect of precursor concentration on stoichiometry and optical properties of spray pyrolyzed nanostructured NiO thin films.

In this study, spray pyrolysis was used to produce nanostructured NiO thin films from high purity nickel acetate (Ni(CH3COO)2.3H2O) precursors on pre-heated ultrasonically cleaned soda-lime glass substrates. The metallic constituent concentrations in the films were varied, and the precursors were produced in distilled water at various molarities ranging from 0.1 to 0.4 M. In the study, the field-emission scanning electron microscope (FESEM) results strongly confirmed adherence of the films to the glass substrate at 350 °C. The presence of Ni and O in the samples was confirmed using Rutherford backscattering spectroscopy (RBS), X-ray diffractometry (XRD) and energy dispersive X-ray spectroscopy (EDX). For the 0.1 M NiO thin films, the thickness was approximately 43 nm, and for the 0.2 M, 0.3 M, and 0.4 M films, the thickness was 46 nm, 47 nm, and 49 nm, respectively. The XRD findings were supported by the increased Raman intensity peaks with increased precursor concentration, which confirmed the films' improved crystallinity. For the same number of passes of films deposition, as the molar concentration increases, the films thickness increases. The amount of nickel in NiO thin films increases as the molarity increases, but the amount of oxygen in NiO thin films decreases as the molarity increases. It was discovered that as molarity increases, the optical transmittance decreases and the optical band gap narrows. The qualities of NiO discovered in this study suggest the films' potentials for usage as window layer and buffer material in thin film solar cells.

Sputum smears quality inspection using an ensemble feature extraction approach.

The diagnosis of tuberculosis (TB) is extremely important. Sputum smear microscopy is thought to be the best method available in terms of accessibility and ease of use in resource-constrained countries. In this paper, research was conducted to evaluate the effectiveness of tuberculosis diagnosis by examining, among other things, the underlying causes of sputum smear quality for Ethiopian states such as Tigray, Amahira, and Oromia. However, because it is done manually, it has its limitations. This study proposes a model for sputum smear quality inspection using an ensemble feature extraction approach. The dataset used was recorded and labeled by experts in a regional lab in Bahir Dar, near Felege Hiwot Hospital after being collected from Gabi Hospital, Felege Hiwot Hospital, Adit Clinic and Gondar Hospital, as well as Kidanemihret Clinic in Gondar. We used a controlled environment to reduce environmental influences and eliminate variation. All the data was collected using a smartphone (the standard 15) with a jpg file extension and a pixel resolution of 1,728 × 3,840. Prior to feature extraction, bicubic resizing, and ROI extraction using thresholding was performed. In addition, sequential Gaussian and Gabor filters were used for noise reduction, augmentation, and CLAHE was used for enhancement. For feature extraction, GLCM from the gray label and CNN from the color image were both chosen. Ultimately, when CNN, SVM, and KNN classifiers were used to test both CNN and GLCM features, KNN outperformed them all with scores of 87, 93, and 94% for GLCM, CNN, and a hybrid of CNN and GLCM, respectively. CNN with GLCM outperformed other methods by 0.7 and 0.1% for GLCM and CNN feature extractors using the same classifier, respectively. In addition, the KNN classifier with the combination of CNN and GLCM as feature extractors performed better than existing methods by 1.48%.

Detection of fake news and hate speech for Ethiopian languages: a systematic review of the approaches.

With the proliferation of social media platforms that provide anonymity, easy access, online community development, and online debate, detecting and tracking hate speech has become a major concern for society, individuals, policymakers, and researchers. Combating hate speech and fake news are the most pressing societal issues. It is difficult to expose false claims before they cause significant harm. Automatic fact or claim verification has recently piqued the interest of various research communities. Despite efforts to use automatic approaches for detection and monitoring, their results are still unsatisfactory, and that requires more research work in the area. Fake news and hate speech messages are any messages on social media platforms that spread negativity in society about sex, caste, religion, politics, race, disability, sexual orientation, and so on. Thus, the type of massage is extremely difficult to detect and combat. This work aims to analyze the optimal approaches for this kind of problem, as well as the relationship between the approaches, dataset type, size, and accuracy. Finally, based on the analysis results of the implemented approaches, deep learning (DL) approaches have been recommended for other Ethiopian languages to increase the performance of all evaluation metrics from different social media platforms. Additionally, as the review results indicate, the combination of DL and machine learning (ML) approaches with a balanced dataset can improve the detection and combating performance of the system.