Ocenjevanje esejev s strojnim učenjem

PERNUŠ, TJAŠA (author), Kononenko, Igor (mentor)

Diplomska naloga se dotika področja strojnega učenja, bolj podrobno pa še področja globokih nevronskih mrež. Cilj naše diplomske naloge je bila primerjava različnih pristopov strojnega učenja pri avtomatskem napovedovanju ocen esejev ter oceniti uspešnost globokih nevronskih mrež v primerjavi z ostalimi modeli. Pri gradnji globokih nevronskih mrež je bila izvedena tudi gradnja globoke nevronske mreže z osnovnimi eseji, ki so razbiti na n-terke, vsaka n-terka pa je predstavljala posamezni atribut. Za primerjavo je bilo uporabljeno okolje R, kjer je bilo izvedeno testiranje in primerjava modelov. Izdelanih je bilo več različnih modelov istega tipa, nato pa za posamezni tip izbran najbolj uspešen, ki je bil nato uporabljen v končni primerjavi različnih tipov modelov.

• Language: Slovenian
• Keywords: ocenjevanje esejev, strojno učenje, globoke nevronske mreže, nevronska mreža, Friedman-Nemenyi test
• Work type: Undergraduate thesis
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2016
• PID: 20.500.12556/RUL-91210
• Publication date in RUL: 24.03.2017

PERNUŠ, TJAŠA, 2016, Ocenjevanje esejev s strojnim učenjem [online]. Bachelor’s thesis. [Accessed 6 April 2025]. Retrieved from: https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=eng&id=91210

Secondary language

• Language: English
• Title: Essay evaluation with machine learning

Abstract

The diploma thesis covers the field of machine learning. In more detail it covers the field of deep neural networks. The goal of our thesis was comparing different approaches of machine learning for building models for automated essay scoring and to evaluate the success of deep neural networks compared to other models. For building models we have used already extracted attributes, but for the deep neural network we have also used original essays, represented by the three attributes, that represent the relationships in a sentence. For comparing we have used the R environment, where we have built, tested and compared the models. Many different models of the same kind were built, from which the best was chosen for further comparison with models of different types.

• Keywords: essay scoring, machine learning, Deep neural network, neural network, Friedman-Nemenyi test