Analiza tekstova projektnih prijedloga primjenom postupaka statističke obrade prirodnoga jezika

Petric Maretić, Hermina (2015) Analiza tekstova projektnih prijedloga primjenom postupaka statističke obrade prirodnoga jezika. Diploma thesis, Faculty of Science > Department of Mathematics.

[img]
Preview
PDF
Language: Croatian

Download (883kB) | Preview

Abstract

This thesis describes an analysis of crowdfunding projects with Natural language processing methods. Projects were downloaded from the two biggest crowdfunding platforms - Kickstarter and Indiegogo and the downloading process has been described in the thesis in short. After that, the preprocessing that transforms raw text into an analysis-friendly form was described. The analysis consists of two parts. The first part is project classification into categories. After the motivation and theoretical background of the SVM method, we have described feature extraction methods and parameters tuning. The classification was then used to classify Indiegogo projects into categories from Kickstarter. This will make further analysis and platforms comparison possible. The second part describes topic modelling on both platforms together. After choosing the number of topics, a model was fitted and the topics were represented by their most frequent words. Finally, the comparison of topics throughout categories connected both parts of the thesis. For the purposes of this thesis, the first section provides methods used for analysis. Additionally, a short description of crowdfunding, as well as an overview of current scientific results.

Item Type: Thesis (Diploma thesis)
Supervisor: Šnajder, Jan and Vuković, Mladen
Date: 2015
Number of Pages: 28
Subjects: NATURAL SCIENCES > Mathematics
Divisions: Faculty of Science > Department of Mathematics
Depositing User: Iva Prah
Date Deposited: 23 Oct 2015 11:35
Last Modified: 23 Oct 2015 11:35
URI: http://digre.pmf.unizg.hr/id/eprint/4181

Actions (login required)

View Item View Item