Recommender system for Slovene electronic books

Lovro Žitnik (2016) Recommender system for Slovene electronic books. EngD thesis.

Preview

Abstract

The thesis describes a prototype of recommender system for Slovenian e-library. Data includes multitude of books in the ePub format, anonymized list of users and a set of transactions between the books and users. Book content was tokenized, POS-tagged and lemmatized. Data extraction methods were used to extract numerical stylometric features of books. Features were evaluated with SPEC and Laplacian scores. We reduced dimensionality of feature vectors of both books and users and clustered them. We used the method proposed by Elkan and Noto as well as one-class SVM method to classify the books. We constructed several variants of recommender systems based on content and collaborative filtering. The evaluation results show that recommender system using only stylometric features is possible, however, collaborative filtering offers better overall performance.

Item Type:

Thesis (EngD thesis)

Keywords:

recommender system, e-books, natural language processing, artificial intelligence, machine learning, clustering, dimensionality reduction, one-class classification

Number of Pages:

100

Language of Content:

Slovenian

Mentor / Comentors:

Name and Surname	ID	Function
izr. prof. dr. Marko Robnik Šikonja	276	Mentor

Link to COBISS:

http://www.cobiss.si/scripts/cobiss?command=search&base=51012&select=(ID=1537048003)

Institution:

University of Ljubljana

Department:

Faculty of Computer and Information Science

Item ID:

3389

Date Deposited:

06 Jul 2016 10:34

Last Modified:

02 Aug 2016 11:04

URI:

http://eprints.fri.uni-lj.si/id/eprint/3389

Actions (login required)

View Item