Institutional Repository of Vilnius University
Matematinio samprotavimo uždavinių tyrimas naudojant mašininio mokymo metodus ir natūralios kalbos apdorojimo algoritmus /
Abstract
dc:descriptionAs artificial intelligence increasingly provides answers to various questions, a fundamental one arises: is finding a solution to all of them is equally simple and easy? Although natural language processing (NLP) methods are capable of analyzing data, they face interpretation challenges in the field of mathematical reasoning problems. To investigate them, during the study, a dataset of mathematical reasoning problems was collected, which, for the purpose of more detailed analysis, was classified into three categories – pattern-based, proof-based and the remaining reasoning tasks that did not belong to either the pattern-based or proof-based categories. To evaluate machine learning capability, two methods were used to translate text into numbers – Bag of Words (BoW) and TF-IDF (Term Frequency-Inverse Document Frequency) along with the algorithms: K-nearest neighbours (KNN), logistic regression and Support Vector Machine (SVM). The results analysis showed that in the task of classifying BoW method is superior to TF-IDF. The highest classification rates were achieved using BoW in combination with the logistic regression algorithm. Due to the particularly distinctive terms used (e.g. prove, justify, etc.) the most easily classified problems are proof type. Meanwhile, non-proof problems are characterized by a wide range of words used, therefore, they may become a significant challenge for some models. The study results revealed that even when training machine learning models, errors cannot be avoided. The principles applied during the study can be used for the systematization of large datasets in any field, especially in the education sector. Such technologies enable the creation of a structured learning enviroment in which students, analogous to machine learning models, develop the ability to learn from abundantly provided data to successfully reach the correct solution.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Petkutė, Ugnė,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD308326903&prefLang=en_US