Search

Due to the widespread distribution of coronavirus and the existence of a massive quantity of data on social networking sites, particularly Twitter, there was an urgent need to develop a model that evaluates users' emotions and determines how they feel about the pandemic. However, the absence of resources to assist Sentiment Analysis (SA) in Arabic hampered the completion of this endeavor. This work presents the ArSentiCOVID lexicon, the first and largest Arabic SA lexicon for COVID-19 that handles negation and emojis. We design a lexicon-based sentiment analyzer tool that depends mainly on the ArSentiCOVID lexicon to perform a three-way classification. Furthermore, we employ the sentiment analyzer to automatically assemble 42K annotated Arabic tweets for COVID-19. We conduct two experiments. First, we test the effect of applying negation and emoji rules to the created lexicon. The results indicate that after applying the emoji, negation, and both rules, the F-score improved by 2.13%, 4.13%, and 6.13%, respectively. Second, we applied an ensemble method that combines four feature groups (n-grams, negation, polarity, and emojis) as input features for eight Machine Learning (ML) classifiers. The results reveal that Random Forest (RF) and Support Vector Machine (SVM) classifiers work best, and that the four feature groups combined are best for representing features produced the maximum accuracy of (92.21%), precision (92.23%), recall (92.21%), and F-score (92.23%) with 3.2% improvement over the base model. [ABSTRACT FROM AUTHOR]

In this paper, a proposed algorithm that dynamically changes the neural network structure is presented. The structure is changed based on some features in the cascade correlation algorithm. Cascade correlation is an important algorithm that is used to solve the actual problem by artificial neural networks as a new architecture and supervised learning algorithm. This process optimizes the architectures of the network which intends to accelerate the learning process and produce better performance in generalization. Many researchers have to date proposed several growing algorithms to optimize the feedforward neural network architectures. The proposed algorithm has been tested on various medical data sets. The results prove that the proposed algorithm is a better method to evaluate the accuracy and flexibility resulting from it. [ABSTRACT FROM AUTHOR]

Dimensionality reduction techniques (DR) enhance data interpretability and reduce space complexity, though at the cost of information loss. Such methods have been prevalent in the Structural Health Monitoring (SHM) anomaly detection literature. While DR is favorable in supervised anomaly detection, where possible novelties are known a priori, the efficacy is less clear in unsupervised detection. In this work, we perform a detailed assessment of the DR performance trade-offs to determine whether the information loss imposed by DR can impact SHM performance for previously unseen novelties. As a basis for our analysis, we rely on an SHM anomaly detection method operating on input signals' fast Fourier transform (FFT). FFT is regarded as a raw, frequency-domain feature that allows studying various DR techniques. We design extensive experiments comparing various DR techniques, including neural autoencoder models, to capture the impact on two SHM benchmark datasets exclusively. Results imply the loss of information to be more detrimental, reducing the novelty detection accuracy by up to 60% with autoencoder-based DR. Regularization can alleviate some of the challenges though unpredictable. Dimensions of substantial vibrational information mostly survive DR; thus, the regularization impact suggests that these dimensions are not reliable damage-sensitive features regarding unseen faults. Consequently, we argue that designing new SHM anomaly detection methods that can work with high-dimensional raw features is a necessary research direction and present open challenges and future directions. [ABSTRACT FROM AUTHOR]

We present empirical findings from using an interactive electronic textbook (eTextbook) system named OpenDSA to teach sophomore- and junior-level Computer Science courses. The web-based eTextbook infrastructure allows us to collect large amounts of data that can provide detailed information about students’ study behavior. In particular we were interested in seeing if the students will attempt to manipulate the electronic resources so as to receive credit without deeply going through the materials. We found that a majority of students do not read the text. On the other hand, we found evidence that students voluntarily complete additional exercises (after obtaining credit for completion) as a study aid prior to exams. We determined that visualization use was fairly high (even when credit for their completion was not offered). Skipping to the end of slideshows was more common when credit for their completion was offered, but also occurred when it was not. We measured the level of use of mobile devices for learning by CS students. Almost all students did not associate their mobile devices with studying. The only time they accessed OpenDSA from a mobile device was for a quick look up, and never for in depth study. [ABSTRACT FROM AUTHOR]