Javascript must be enabled for the correct page display

iNewsReader: A Personal Netnews recommender

Groot, H.T. de (2011) iNewsReader: A Personal Netnews recommender. Master's Thesis / Essay, Artificial Intelligence.

[img]
Preview
Text
AI-MAI-2011-H.T.de.Groot.pdf - Published Version

Download (5MB) | Preview

Abstract

The Internet: Web sites; twitter; personal messages; blogs; RSS feeds... Continuously, an overwhelming growing amount of information reaches our human brain. Mostly only a small part of this information is of true personal interest and therefore tracking the interesting part becomes harder every day. This is referred to as the problem of information overflow. Approaches to tackle this problem are the usage of news summaries, community interests and collaborative intelligence algorithms. In this research a solution is sought by developing a personalized adaptive netnews recommender system: iNewsReader. In general, recommender systems try to deliver personalized items or information of interest based on a profile of the user. In this research an advanced framework design is proposed based on multi-agent technologies and known working technologies from literature. As a proof of concept the main modules (web crawler, data storage, portal & recommenders) and the crucial parts (agents) of this framework are implemented for research. The focus of this recommender system lies on the content-based recommendation methods, which are mainly based on Support Vector Machine and Naive Bayes machine learning algorithms for text classification. The implemented recommender system is accessible through a web-portal and the performance is tested in a small experiment on continuous real time data from the internet. During this experiment multiple pre-configured agents try to collect articles of personal interest by creating user models from the users' feedback and browsing history. Using these personal user models, agents collect news article data from a growing multi-dimensional data storage. Next, the selected articles are presented to the above classification algorithms to form the final personal recommendation. The results of the conducted experiment show a positive effect on learning and recommendation performance, but additions and improvements on many parts of the system can probably elevate this result enormously.

Item Type: Thesis (Master's Thesis / Essay)
Degree programme: Artificial Intelligence
Thesis type: Master's Thesis / Essay
Language: English
Date Deposited: 15 Feb 2018 07:45
Last Modified: 15 Feb 2018 07:45
URI: http://fse.studenttheses.ub.rug.nl/id/eprint/9531

Actions (login required)

View Item View Item