Marcelo do Vale Cunha

COMPLEX SEMANTIC NETWORKS

Verbal language is a dynamic mental process. Ideas emerge by means of the selection of words from subjective and individual characteristics throughout the oral discourse. The goal of this work is to characterize the complex network of word associations that emerge from an oral discourse from a discourse topic. Because of that, concepts of associative incidence and fidelity have been elaborated and represented the probability of occurrence of pairs of words in the same sentence in the whole oral discourse. Semantic network of words associations were constructed, where the words are represented as nodes and the edges are created when the incidence-fidelity index between pairs of words exceeds a numerical limit (0.001). Twelve oral discourses were studied. The networks generated from these oral discourses present a typical behavior of complex networks and their indices were calculated and their topologies characterized. The indices of these networks obtained from each incidence-fidelity limit exhibit a critical value in which the semantic network has maximum conceptual information and minimum residual associations. Semantic networks generated by this incidence-fidelity limit depict a pattern of hierarchical classes that represent the different contexts used in the oral discourse.

Instrumento livre para medidas de movimento

This paper presents the validation of a computational tool that serves to obtain continuous measurements of moving objects. The software uses techniques of computer vision, pattern recognition and optical flow, to enable the tracking of objects in videos, generating data trajectory, velocity, acceleration and angular movement. The program was applied to track a ball around a simple pendulum. The methodology used for this validation takes as a basis for comparisons the measured values of the program as well as the theoretical values expected according to the model of a simple pendulum. The experiment is appropriate to the method because it was built within the limits of the linear harmonic oscillator, and energy losses due to friction had been minimized, making it the most ideal possible. The results indicate that the tool is sensitive and accurate. Deviations of less than a millimeter in the measurements of the trajectory ensure the applicability of the software on physics, either in research or in teaching topics.