A chatbot or conversational agent is a software that can interact or ``chat'' with a human user using a natural language, like English, for instance. Since the first chatbot developed, many have been created but most of their problems still persist, like providing the right answer to the user and user acceptance itself. Considering such facts, in this work, we present a chatbot-building framework that considers the use of sentiment analysis and tree timelines to provide a better chatbot answer. For instance, as presented in our experiments, the user can be addressed to a human attendant when its sentiment is very negative, or even try another branch of the tree timeline, as an alternative answer, whenever the user sentiment is less negative.
Multi-objective decision-making in multi-agent scenarios poses multiple challenges. Dealing with multiple objectives and non-stationarity caused by simultaneous learning are only two of them, which have been addressed separately. In this work, reinforcement learning algorithms that tackle both issues together are proposed and applied to a route choice problem, where drivers must select an action in a single-state formulation, while aiming to minimize both their travel time and toll. Hence, we deal with repeated games, now with a multi-objective approach. Advantages, limitations and differences of these algorithms are discussed. Our results show that the proposed algorithms for action selection using reinforcement learning deal with non-stationarity and multiple objectives, while providing alternative solutions to those of centralized methods.
Three-phase induction motors are widely used in different applications in the industry due to their robustness, low cost, and reliability. Untimely identification and correct diagnosis of incipient faults reduce cost and improve the maintenance management of these machines. This paper explores a new method for robust classification of rotor failures in three-phase induction motors (MITs) connected directly to the electrical network, operating in a steady-state, under unbalanced voltages and load conditions. Through an innovative methodology, an analysis of the electrical current signals from 1 hp and 2 hp motors in the frequency domain was performed. Such analysis was applied in constructing input matrices for a Multilayer Perceptron Neural Network (MLPNN) to detect faults. Furthermore, this methodology proved to be robust because the samples of the failing and healthy motors include voltage unbalance conditions in the electrical supply and a significant variation in the load applied to the motor shaft. Such load variation was used for the detection of failures of 1, 2, and 4 broken bars consecutively on the rotor and in the condition of 2 broken bars and 2 other broken bars diametrically opposite. The results were promising and were obtained using 847 real samples from an experimental bench used to construct the neural model and its respective validation.
The World Health Organization (WHO) has declared the novel coronavirus (COVID-19) outbreak a global pandemic in March 2020. Through a lot of cooperation and the effort of scientists, several vaccines have been created. However, there is no guarantee that the virus will shortly disappear, even if a large part of the population is vaccinated. Therefore, non-invasive methods, with low cost and real-time results, are important to detect infected individuals and enable earlier adequate treatment, in addition to preventing the spread of the virus. An alternative is using forced cough sounds and medical information to distinguish a healthy person from those infected with COVID-19 via artificial intelligence. An additional challenge is the unbalancing of these data, as there are more samples of healthy individuals than contaminated ones. We propose here a Deep Neural Network model to classify people as healthy or sick concerning COVID-19. We used here a model composed by an Convolutional Neural Network and two other Neural Networks with two full-connected layers, each one trained with different data from the same individual. To evaluate the performance of the proposed method, we combined two datasets from the literature: COUGHVID and Coswara. That dataset contains clinical information regarding previous respiratory conditions, symptoms (fever or muscle pain), and a cough record. The results show that our model is simpler (with fewer parameters) than those from the literature and generalizes better the prediction of infected individuals. The proposal presents an average Area Under the ROC Curve (AUC) equal to 0.885 with a confidence interval (0.881 - 0.888), while the literature reports 0.771 with (0.752 - 0.783).
This paper presents the performance evaluation of a wireless sensor mesh network, for monitoring temperature and humidity in horticultural products when transported in truck galleys. For this purpose a software solution was proposed using ESP8266 devices powered by batteries. The mesh network was managed by the painlessMesh library. The proposed solution aims to minimize the energy consumption of the sensor nodes. The validation of the solution was performed in an area simulating a galley of a truck, where five sensor nodes and a root node were distributed. The tests were developed considering four different models involving variations in messages delivery confirmation, number of attempts until successful delivery and duty cycle duration of the nodes. The performance evaluation of the solution aimed to determine, connectivity rate, sending rate after connection and delivery rates of the first and second attempts. The results obtained show that the message delivery confirmation does not bring added value to the solution, contributing only to increase energy consumption. The use of synchronous duty cycles also showed worse results than the asynchronous use. These results allow the creation of a knowledge base for the use of this solution in a real context.
