| Studiju veids |
bakalaura akadēmiskās studijas |
| Studiju programmas nosaukums |
Datorsistēmas |
| Nosaukums |
Noskaņojuma analīze Steam platformas datorspēļu atsauksmēs |
| Nosaukums angļu valodā |
Sentiment Analysis of Steam Game Reviews |
| Struktūrvienība |
33000 Datorzinātnes, informācijas tehnoloģijas un enerģētikas fakultāte |
| Darba vadītājs |
Gints Jēkabsons |
| Recenzents |
Lektors Māra Pudāne |
| Anotācija |
The bachelor thesis presents a study of Sentiment Analysis for video game reviews of the Steam platform. The goal of the thesis is to compare effectiveness of different classification algorithms in Sentiment analysis of Steam game reviews, develop software for sentiment prediction, and implement text highlighting for explanation of predicted class. The thesis examines concepts that affect this topic such as Natural language Processing, Pre-processing techniques, algorithms in Machine learning, TF-IDF. Taking into account these elements, a software is developed for the purpose of classifying text and predicting the sentiment label of reviews. The software is then used to perform experiments with different parameters (pre-processing steps, dataset size), classification algorithms and gather results selecting the best classification algorithm, and visualize how words of a review are influencing the predicted sentiment label.
The bachelor thesis contains 55 pages, 17 images, 22 tables, 1 Appendix and 74 reference sources. |
| Atslēgas vārdi |
NOSKAŅOJUMA ANALĪZE, MAŠĪNMĀCĪŠANĀS ALGORITMI, DABISKĀS VALODAS APSTRĀDE, KLASIFICĒSANA |
| Atslēgas vārdi angļu valodā |
SENTIMENT ANALYSIS, MACHINE LEARNING ALGORITHMS, NATURAL LANGUAGE PROCESSING, CLASSIFICATION MODEL |
| Valoda |
eng |
| Gads |
2020 |
| Darba augšupielādes datums un laiks |
05.06.2020 08:53:57 |
