نشریه محاسبات نرم
سال یازدهم شماره 1 (پیاپی 21، بهار و تابستان 1401)

Nowadays, according to the growing volume of electronic documents, the classification of text has attracted the attention of information retrieval researchers. Considering the importance of text classification and the efforts done in this field in several languages in the world, the necessity of Persian text classification is understood. In general, we can classify text classification methods into two classes, including traditional methods (based on feature selection and machine learning) and methods based on deep learning. Deep learning methods, due to the ability of weight sharing, significantly reduce the number of trainable parameters and thus increase generalization and provide better results than other methods. There are a few methods based on deep learning for Persian text classification. In this study, we propose to use CNN and BLSTM with an attention layer for Persian text classification named ParsCNN and ParsBiLSTM. The experimental results on the Hamshahri dataset show that the ParsCNN method has a precision of 0.69, a recall of 0.7, and, an F-score of 0.69; Also, the ParsBiLSTM method has a precision of 0.72, a recall of 0.73 and, an F-score of 0.72, which indicates the methods based on deep Neural Networks have better performance than other approaches.

In this paper, the rules governing the behavior of the system are extracted from the system specifications and then they are modeled by Petri-nets. The set of these rules forms knowledge base of the system, which provides the ability of making inferences. Each rule consists of several premises and a conclusion. When there are many rules and the premises in different rules overlap, it is possible to prevent the repetition of premises using a hierarchical structure by the inference engine and thus reduce the number of checks required to reach the conclusion. When these rules have many and fuzzy variables, they take a complex form, and it becomes difficult to understand and deduce their behavior. To better understand this complexity, it is appropriate to visualize it using fuzzy petri-nets. So far, many different methods based on fuzzy Petri-nets have been presented to model fuzzy rules. But these methods either do not support the large number of rules and variables or do not consider matters like the role of conditional propositions in the occurrence of the conclusion propositions, the probability of the conclusion propositions, the threshold value for the conditional propositions and the conclusions, the certainty factor for the rule or for the conditional propositions. In this paper, by extending our previous work, we present a model based on fuzzy Petri-nets that covers the two mentioned cases. Finally, we present the proposed model for a secure water refinement system and the attacks against it.

In this paper, by considering the notions of maximal, prime, and minimal prime filters in residual lattices, we study their properties and some examples of residuated lattices are given. It is reminded first the notion of a filter of a residuated lattice and then we show that each proper filter is contained in a maximal filter and therefore in a prime filter. We investigate the prime filters in an MTL algebra and state and prove the fundamental theorem of the prime filters for residuated lattices. We investigate the residuated lattices in which each its filter is principal, and by using the prime filters, we inspect co-annihilators. Then the notion of a minimal prime filter in a residuated lattice is introduced and a fundamental theorem for minimal prime filters in residuated lattice is stated and proved. Finally, we introduce the divisor filters in a residuated lattice and by using them; we investigate the minimal prime filters

In this paper, we present an efficient numerical method for first order nonlinear singularly perturbed differential equations. This method is based on a hybrid block method with four hybrid points on a non-uniform mesh. The main characteristics of the hybrid block method, including consistency, zero stability and stability region are investigated. In order to optimize mesh points, we will use the interpolation technique and arc-length monitor function. After finding the new optimized mesh points, we can apply the proposed hybrid block method to optimize the numerical solution. The numerical experiments show the efficiency and accuracy of the proposed method.

Human activity recognition can be employed in systems that provide support to people who need special care in their daily activities. Various machine learning methods, including Hidden Markov Models and their extensions, are widely used to deal with this problem. In the previous works, HMM-based methods use the conditional independent assumption to compute the probability of a segment of observations. In this research, instead of conditional independent assumption, a new probabilistic model for string space, based on Dynamic Time Warping and Weighted Levenstein Distance is proposed. The model, combined with the Hidden semi Markov Model, is applied to a publicly available dataset. The results show considerable improvements in comparison with using the Hidden semi Markov Model independently. The proposed models are flexible and can work together with other probabilistic models.

In the last decade, with the expansion of the World Wide Web, the speed and ease of access to ideas, documents, articles, manuscripts, and data collected by others has increased. This has made the exchange of information and ideas between researchers and producers of science easier, but on the other hand, it has made it easier to apply unauthorized copies, write summaries without mentioning the source, and steal literary texts in general. Since universities and educational centers make scientific and research resources easily available to most users, recognizing the authenticity of scientific texts in these centers is more important and, of course, more sensitive. In this research, a method is presented to compare the related parts using the blocking of document parts. In the proposed method, after classifying the documents into two categories of main documents and suspicious documents, preprocessing has been done with the aim of eliminating word stops and new wording. Then the documents are segmented and using cosine similarity, the degree of similarity of the texts with each other is determined. The proposed method in the test of 50 documents in the data set has an accuracy of 94%, which is an improvement of 2% compared to one of the similar methods.

In most applications and scientific articles, for simplicity the games have been done as one shot. Also, in most of the games, the mixed strategies are not used, whereas a mix of two or more strategies can be more helpful for players in future. Therefore, it is quite natural considering a mix of two or more strategies stochastically. This article introduces a series of games called Markovian Dynamic Games that if you consider the strategies of each player as a state, the player can select the same or another state depending on the situation in the next steps. The selecting each state in each step will be done with a specified probability.This probability is depending on the strategies of the competing players. In this study, a two-player discrete game with Markov chain approach is considered that the probability of transfer is well-known, independent and is only influenced by the competitorchr(chr('39')39chr('39'))s previous strategies. In this paper, the equilibrium points in markovian dynamic games are evaluated and analyzed. Numerical examples are also presented for more explanations. In this paper, the difference between single stage and multi-stage games in determining equilibrium points is shown. If a game is played in a multilevel manner, it is possible to design a discrete game as a Markov chain using the probability of transferring and considering the strategies of the game as a mode in each step, and by determining the probabilities of the specified chain, points He gained some balance.For this purpose, dynamic games were considered,

Today, the Internet, especially social networks such as Twitter, Facebook, and Telegram, has become a platform for exchanging ideas and sharing user opinions. Sentiment analysis based on user opinions in these networks can help explain and predict social phenomena and find suitable products or services for individuals, companies, and organizations. So far, a lot of research has been done on social media data in English; But limited research has been done for the Persian language. In this paper, a Sentiment analysis system on Persian Telegram data is proposed. For this purpose, several feature extraction methods including Countvectorizer, TF-IDF, and word embedding matrix have been studied to represent textual data numerically. Then, to classify the data, different classical machine learning methods including support vector machine, decision tree, K-nearest neighbor, Naïve Bayes, and logistic regression, the combination of classical methods as well as deep learning methods including deep neural network (DNN), convolutional neural network (CNN), long short-term memory network and bidirectional long short-term memory network has been investigated. Finally, the evaluation and analysis of the results on the data collected from Persian Telegram shows that the best performance has been obtained by word embedding and bidirectional long short-term memory network with an accuracy of 90.67%, precision of 90.01%, recall of 89.54% and F1 of 89.77%.