Vietnam Journal of Computer Science

Ontologies have developed into a prevailing technique for establishing semantic interoperability among heterogeneous systems transacting information. An ontology is an unambiguous blueprint of a concept. For Artificial Intelligence, only the defined notions can be considered existent. Thus, in relation to AI, an ontology can be understood as part of a program which delineates a collection of descriptions. An ontology, therefore, correlates the labels of the entities in the universe of discourse with wording that holds meaning for humans, explaining what those labels signify, along with the precise principles that force the interpretation and semantic utilization of these labels. An ontology constitutes a proper statement of a logical theory. It is a crucial component of a system with the capability to process, manage, analyze, correlate and reason from the large datasets characterized by heterogeneity. This paper depicts the process of development of a Common Representational Model (CRM) on top of several ontologies, taxonomies and classifications to facilitate computational and data mining functionalities. The building blocks of said CRM are delineated in detail, as well as its application in a specific use case.

In this paper, we have explored different strategies to find out the saliency of the features from different pretrained models in detecting violence in videos. A dataset has been created which consists of violent and non-violent videos of different settings. Three ImageNet models; VGG16, VGG19, ResNet50 are being used to extract features from the frames of the videos. In one of the experiments, the extracted features have been feed into a fully connected network which detects violence in frame level. Moreover, in another experiment, we have fed the extracted features of 30 frames to a long short-term memory (LSTM) network at a time. Furthermore, we have applied attention to the features extracted from the frames through spatial transformer network which also enables transformations like rotation, translation and scale. Along with these models, we have designed a custom convolutional neural network (CNN) as a feature extractor and a pretrained model which is initially trained on a movie violence dataset. In the end, the features extracted from the ResNet50 pretrained model proved to be more salient towards detecting violence. These ResNet50 features, in combination with LSTM provide an accuracy of 97.06% which is better than the other models we have experimented with.

Many present software systems can be developed by a sequence of transformations from the source specification to the final implementation. An interesting question is whether we can support such a sequence of transformations by some formal apparatus that enables to verify succeeding steps of development, and finally also the whole development process. As an example, we use the transformation of a definition of the set of system nodes defined as classical workflow models, and then transform them into a set of Petri nets representing the target system implementation. Such a transformation supports development of software systems, whose specification is based on classical workflow models, but the implementation is based on Petri nets. Each part of the designed system is translated from workflow model into a set of Petri nets, and interpreted by the set of special Petri Nets Virtual Machines (PNVMs) which are installed on all nodes of the system. The method is illustrated on the example of house heating system.

In network telemetry systems, nodes produce vast number of configuration files based on how they are configured. Steps were taken to process these files into databases to help the work of the developers, testers and customer support to focus on the development and testing and to be able to give advice to the customers about how to configure the nodes. However, the processing of these data in relational database manager system is slow, hard to query and the storage takes huge disk space.

In this paper, we are presenting a way to store the data produced by these nodes in graph database, changing from relational database to NoSQL environment. With our approach, one can easily represent and visualize the network of machines.

In the end, we are going to compare the inserting, querying time and storage size in different database manager systems. The results could also be used for other types of configuration data too from other kinds of machines to show the connection between them and query them easily.

Recently, the advances in communication technologies have made music retrieval easier. Without downloading the music, the users can listen to music through online music websites. This incurs a challenging issue of how to provide the users with an effective online listening service. Although a number of past studies paid attention to this issue, the problems of new user, new item and rating sparsity are not easy to solve. To deal with these problems, in this paper, we propose a novel music recommender system that fuses user contents, music contents and preference ratings to enhance the music recommendation. For dealing with problem of new user, the user similarities are calculated by user profiles instead of traditional ratings. By the user similarities, the unknown ratings can be predicted using user-based Collaborative Filtering (CF). For dealing with problems of rating sparsity and new items, the unknown ratings are initialized by acoustic features and music genre ratings. Because the unknown ratings are initially imputed, the rating data will be enriched. Thereupon, the user preference can be predicted effectively by item-based CF. The evaluation results show that our proposed music recommender system performs better than the state-of-the-arts methods in terms of Root Mean Squared Error.

Predicting the solutions of complex systems is a crucial challenge. Complexity exists because of the uncertainty as well as nonlinearity. The nonlinearity in complex systems makes uncertainty irreducible in several cases. In this paper, two new approaches based on neural networks are proposed in order to find the estimated solutions of the fully fuzzy nonlinear system (FFNS). For obtaining the estimated solutions, a gradient descent algorithm is proposed in order to train the proposed networks. An example is proposed in order to show the efficiency of the considered approaches.