مجله رایانش نرم و فناوری اطلاعات
سال سیزدهم شماره 1 (بهار 1403)

Today, one of the most popular and challenging tasks in natural language processing is answering complex user questions. Question-answering systems, as a new generation of search engines, receive user questions in natural language without semantic limitations and provide precise answers. In recent years, most research in the field of question-answering systems has been focused on the English language, and not much effort has been made in languages with limited resources, such as Persian. This limitation prevents these systems from efficiently handling questions in languages like Persian.In this regard, this article aims to enhance the efficiency of question-answering systems in the Persian language by creating a dataset for answering complex multi-turn questions. Multi-hop questions, require at least two steps of reasoning to reach an answer. This dataset, called PersianMHQA, is the first open-domain question-answering dataset in Persian and includes 7,000 multi-hop questions. It was generated using the Persian Wikipedia as a knowledge source. To evaluate and benchmark this dataset, it has been fine-tuned on the latest pre-trained language models that support the Persian language.The best results obtained based on F1 score and exact match on this dataset are 75.92% and 71.73%, respectively. These results indicate that this dataset is a powerful starting point for improving multi-hop complex question-answering for Persian language systems.In this regard, this article aims to enhance the efficiency of question-answering systems in the Persian language by creating a dataset for answering complex multi-turn questions. Multi-hop questions, require at least two steps of reasoning to reach an answer. This dataset, called PersianMHQA, is the first open-domain question-answering dataset in Persian and includes 7,000 multi-hop questions. It was generated using the Persian Wikipedia as a knowledge source. To evaluate and benchmark this dataset, it has been fine-tuned on the latest pre-trained language models that support the Persian language.The best results obtained based on F1 score and exact match on this dataset are 75.92% and 71.73%, respectively. These results indicate that this dataset is a powerful starting point for improving multi-hop complex question-answering for Persian language systems.

Social Internet of Things (SIoT) is an emerging research field that originates from the combination of Internet of Things (IoT) and social networks. With the large number of users and smart objects and the high volume of relationships that exist in the social Internet of Things, a recommendation system is very necessary to help users find the smart objects they need or even suitable users as friends or future colleagues. In addition, SIoT has the potential to create new opportunities for recommender systems, as it can use social data to provide more relevant and personalized recommendations for recommending services in the Internet of Things. On the other hand, in recommendation systems in social networks, the use of data available in the Internet of Things makes suggestions more accurate. In this article, a recommendation system for the social Internet of Things is designed. The first phase of this proposed method obtains the similarity of users in the Internet of Things by using the preference of users to use smart devices plus the time and place of using the smart device. In the second phase, relationship similarity and profile similarity are calculated. Finally, by obtaining the sum of existing similarities, k-top users with the highest similarity with the target user are recommended to find future friends and colleagues. In this proposed method, the evaluation criteria have improved by about 5% compared to similar methods, and due to the similarity of the profile, it is also effective for improving cold start. In addition, by removing insecure users from the proposed set, the security of this proposed system is provided to a significant extent.

With the advent of the next generation of wireless communication networks, the requirements for high-rate and timely information delivery, especially in real-time applications, have increased dramatically. In particular, Millimeter wave (mmWave) communications can provide a high data rate, and reconfigurable intelligent surfaces (RISs) can reduce the blockage sensitivity of mmWave links. Also, to quantify the data freshness, the age of information (AoI) metric is introduced in the recent literature, defined as the time elapsed since the last successfully received status update was generated. This paper aims to optimize the average AoI in a multi-user RIS-assisted millimeter wave network. For this purpose, the proposed problem was formulated as a Markov decision process (MDP) that aims to obtain the optimal control policy that minimizes the sum of the average AoI of all users. Under the unknown system statistics assumption, we propose a model-free deep reinforcement learning algorithm to schedule the users’ transmission, adjust the transmit power, and configure the RIS to reflect the signals.  The performance of the proposed method is evaluated in terms of convergence characteristics and the impact of changes in network parameters.

Early detection of plant diseases is crucial for timely prevention of pest infestation. In this paper plant diseases are accurately identified through image processing of leaves and extracting various features such as color, texture, and shape. Generally, plant diseases cause significant changes in the color and appearance of leaves. The color feature can be used to determine the type of plant disease. This paper employs moment and Hu moment features for extracting color characteristics from leaves, and area and perimeter features for detecting shape changes. To enhance the proposed features, additional texture features such as Local Binary Pattern (LBP), co-occurrence matrix, and vein patterns are utilized. Finally, to prevent model complexity and overfitting, the mRMR feature selection method is applied to eliminate irrelevant features. After developing the final model, healthy and diseased samples are classified using Random Forest and Support Vector Machine (SVM) classifiers. The proposed method has been evaluated on the PlantVillage dataset with 87867 samples. The average accuracy of the proposed method in diagnosing diseases in 9 different plant samples using Random Forest and SVM classifiers is 97.63% and 94.98%, respectively, which surpasses existing methods.

Nowadays, short message communication has a wide range of applications in modern life. Authentication, announcements, reports, and personal communications are the common instances of application of this widely used digital media. At transmission time, the message content spreads in the air without any intrinsic security countermeasure. As a result, the content of messages is interceptable without the need for complicated equipment. In modern methods, security has migrated from hardware to software, especially in smartphones. In this research, a method for securing short message communications is introduced, which uses four distinguished key generation factors to produce a couple of unique and dynamic encryption and decryption keys. The evaluation results show that the proposed method uses secure encryption and decryption keys, is resistant against a wide range of cryptographic and network attacks, outstandingly reduces the cost of establishing secure sessions, and highly satisfies the users with the delivered level of security and usability.

Progressive Web Apps (PWA) are a new web development paradigm that is being used in a variety of web apps. With this new paradigm, web apps can provide features similar to native apps, blurring the line between websites and native apps. However, new challenges may arise during the development of PWAs. Understanding the issues that developers may encounter while developing PWAs is necessary due to their widespread interest. These issues have not yet been extensively identified and analyzed in the existing research. Therefore, this research aims to fill this knowledge gap by conducting an empirical study of PWA-related questions on Stack Overflow (SO), a well-known website for developers. Our process involves obtaining a list of PWA tags and extracting 8,654 questions related to PWAs from SO. To comprehend the issues being discussed by PWA developers on SO, we use topic modeling. After that, we analyze their characteristics (such as the challenges’ popularity and difficulty). Our research shows that developers encounter eleven challenges, including the Service Worker lifecycle. Our study deepens our understanding of PWA development issues and provides recommendations for researchers, educators, and developers to promote PWA adoption.