نشریه مهندسی برق و مهندسی کامپیوتر ایران
سال نوزدهم شماره 3 (پیاپی 65، پاییز 1400)

Opinion mining as a sub domain of data mining is highly dependent on natural language processing filed. Due to the emerging role of e-commerce, opinion mining becomes one of the interesting fields of study in information retrieval scope. This domain focuses on various sub areas such as polarity detection, aspect elicitation and spam opinion detection. Although there is an internal dependency among these sub sets, but designing a thorough framework including all of the mentioned areas is a highly demanding and challenging task. Most of the literatures in this area have been conducted on English language and focused on one orbit with a binary outcome for polarity detection. Although the employment of supervised learning approaches is among the common utilizations in this area, but the application of deep neural networks has been concentrated with various objectives in recent years so far. Since the absence of a trustworthy and a complete framework with special focuses on each impacting sub domains is highly observed in opinion mining, hence this paper concentrates on this matter. So, through the usage of opinion mining and natural language processing approaches on Persian language, the deep neural network-based framework called RSAD that was previously suggested and developed by the authors of this paper is optimized here to include the binary and numeric polarity detection output of sentences on aspect level. Our evaluation on RSAD performance in comparison with other approaches proves its robustness.

Nowadays, office, residential, and historic buildings often require special monitoring. Obviously, such monitoring involves costs, errors and challenges. As a result of factors such as lower cost, broader application, and ease of installation, wireless sensor networks are frequently replacing wired sensor networks for structural health monitoring. Depending on the type and condition of a structure, factors such as energy consumption and accuracy, as well as fault tolerance are important. Particularly when wireless sensor networks are involved, these are ongoing challenges which, despite research, have the possibility of being improved. Using the Markov decision process and wake-up sensors, this paper proposes an innovative approach to monitoring stable and semi-stable structures, reducing the associated cost and error over existing methods, and according to the problem, we have advantages both in implementation and execution. Thus, the proposed method uses the Markov decision process and wake-up sensors to provide a new and more efficient technique than existing methods in order to monitor the health of stable and semi-stable structures. This approach is described in six steps and compared to widely used methods, which were tested and simulated in CupCarbon simulation environment with different metrics, and shows that the proposed solution is better than similar solutions in terms of a reduction of energy consumption from 11 to 70%, fault tolerance in the transferring of messages from 10 to 80%, and a reduction of cost from 93 to 97%.

Support Vector machine is one of the most popular and efficient algorithms in machine learning. There are several versions of this algorithm, the latest of which is the fuzzy least squares twin support vector machines. On the other hand, in many machine learning applications input data is continuously generated, which has made many traditional algorithms inefficient to deal with them. In this paper, for the first time, an incremental version of the fuzzy least squares twin support vector algorithm is presented. The proposed algorithmis represented in both online and quasi-online modes. To evaluate the accuracy and precision of the proposed algorithmfirst we run our algorithm on 6 datasets of the UCI repository. Results showthe proposed algorithm is more efficient than other algorithms (even non-incremental versions). In the second phase in the experiments, we consider an application of Internet of Things, and in particular in data related to daily activities which inherently are incremental. According to experimental results, the proposed algorithm has the best performance compared to other incremental algorithms.

With the rapid development of new technologies in the field of internet of things (IoT) and intelligent networks, researchers are more interested than ever in the concept of wireless sensor networks (WSNs). The emergence of these densely structured networks in recent years has raised the importance of the use of telecommunications technologies, such as ultra-wideband (UWB) technology with high reliability, industrial applications, and appropriate communication security. However, there are still numerous concerns about the extent of inter-network interference, particularly owing to undesired spectral discrete lines in this technology. As a result, it is necessary to provide an optimal solution to eliminate interference and control the power spectrum, and then design the optimal transmitter-receiver structures while considering high sensitivities to the synchronization problem in WSNs based on UWB technology. These goals are pursued in the present study by employing the optimal spectral strategy in the signal model, the structure of the transmitter sensor, and then constructing the optimal or sub-optimal receiver sensor structures, the results of which indicate improved communication performance in WSNs.

The efficiency of human action recognition systems depends on extracting appropriate representations from the video data. In recent years, deep learning methods have been proposed to extract efficient spatial-temporal representations. Deep learning methods, on the other hand, have a high computational complexity for development over temporal domain. Challenges such as the sparsity and limitation of discriminative data, and highly noise factors increase the computational complexity of representing human actions. Therefore, creating a high accurate representation requires a very high computational cost. In this paper, spatial and temporal deep learning networks have been enhanced by adding appropriate feature selection mechanisms to reduce the search space. In this regard, non-online and online feature selection mechanisms have been studied to identify human actions with less computational complexity and higher accuracy. The results showed that the non-linear feature selection mechanism leads to a significant reduction in computational complexity and the online feature selection mechanism increases the accuracy while controlling the computational complexity.

The efficiency of human action recognition systems depends on extracting appropriate representations from the video data. In recent years, deep learning methods have been proposed to extract efficient spatial-temporal representations. Deep learning methods, on the other hand, have a high computational complexity for development over temporal domain. Challenges such as the sparsity and limitation of discriminative data, and highly noise factors increase the computational complexity of representing human actions. Therefore, creating a high accurate representation requires a very high computational cost. In this paper, spatial and temporal deep learning networks have been enhanced by adding appropriate feature selection mechanisms to reduce the search space. In this regard, non-online and online feature selection mechanisms have been studied to identify human actions with less computational complexity and higher accuracy. The results showed that the non-linear feature selection mechanism leads to a significant reduction in computational complexity and the online feature selection mechanism increases the accuracy while controlling the computational complexity.

Analysis of published news content is one of the most important issues in information retrieval. Much research has been conducted to analyze individual news articles, while most news events in the media are published in the form of several related articles. Event detection is the task of discovering and grouping documents that describe the same event. It also facilitates better navigation of users in news spaces by presenting an understandable structure of news events. With rapid and increasing growth of online news, the need for search engines to retrieve news events is felt more than ever. The main assumption of event detection is that the words associated with an event appear in the same time windows and similar documents. Accordingly, in this research, we propose a retrospective and feature-pivot method that clusters words into groups according to semantic and temporal features. We then use these words to produce a time frame and a human readable text description. The proposed method is evaluated on the All The News dataset, which consists of two hundred thousand articles from 15 news sources in 2016 and compared to other methods. The evaluation shows that the proposed method outperforms previous methods in terms of precision and recall.