Open AccessSEMANTIC NETWORK TRANSFORMATION METHOD FOR AUTOMATION OF PROGRAMMING PROBLEMS SOLUTIONS EVALUATION IN E-LEARNING
Author(s) -
А. В. Федоров,
Alexey Shikov
Publication year - 2020
Publication title -
vestnik astrahanskogo gosudarstvennogo tehničeskogo universiteta. seriâ: upravlenie, vyčislitelʹnaâ tehnika i informatika
Language(s) - English
Resource type - Journals
eISSN - 2224-9761
pISSN - 2072-9502
DOI - 10.24143/2072-9502-2020-4-7-17
Subject(s) - computer science , perceptron , transformation (genetics) , mnist database , automation , algorithm , basis (linear algebra) , artificial intelligence , process (computing) , machine learning , data mining , artificial neural network , programming language , mathematics , mechanical engineering , biochemistry , chemistry , geometry , engineering , gene
The article presents a semantic network transformation method for a programcode into an N-dimensional vector. The proposed method allows automating the quality assessment of solving programming problems in the process of e-learning. The method includes the authentic algorithms of building and converting the network. In order to determine the algorithm in the program code there is a template of this algorithm, presented in the form of a subgraph of abstract concepts of the language in the semantic network, built on the basis of this code. The search for the algorithm by comparing the subgraph of the network with the template network helped to identify the BFS algorithm with a given accuracy: the cutoff threshold for the perceptron outputs is 0.85, which is based on the calculation of accuracy of the single-layer perceptron in the classification of the MNIST base equal to 88%, which confirms the effectiveness of the developed method and requires further research using machine learning methods to find the optimal value of the coordinates of the nodes of the semantic network and templates of algorithms.