04.11.2014 Views

elektronická verzia publikácie - FIIT STU - Slovenská technická ...

elektronická verzia publikácie - FIIT STU - Slovenská technická ...

elektronická verzia publikácie - FIIT STU - Slovenská technická ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

User Modeling for Personalized Web-Based Systems 243<br />

Web<br />

Crawler<br />

Classifier<br />

Paper<br />

extractor<br />

Author<br />

extractor<br />

Conference<br />

extractor<br />

Location<br />

extractor<br />

Paper<br />

aggregator<br />

Author<br />

aggregator<br />

Conference<br />

aggregator<br />

Location<br />

aggregator<br />

Scientific<br />

Object<br />

warehous<br />

Figure 8-9. Architecture of Scientific Papers Extraction System (according to [50]).<br />

good results in scientific papers domain. If applied to multiple domain, the system could<br />

possibly provide enough information about user to overcome the cold-start problem.<br />

Social Networks Discovering The paper [55] proposes the idea, that each recommender<br />

system naturally fosters communities of users. It is exactly the community which drives the<br />

recommendation once enough users are engaged and modeled (see Figure 8-10).<br />

However, as an implicit user modeling does not make the social network identification<br />

salient, authors propose to employ other techniques to discover social networks:<br />

– Link Analysis and Cyber-Communities – evidence of community existence is often implicit<br />

in data, such as communications logs and webpages. These are fertile reflections<br />

of natural connectivity among people. Some recommender systems require users to<br />

create and maintain profiles. On the other hand, approaches which model people connections<br />

or social organization result in representations which are likely to be more<br />

accurate reflections than a user’s perception of his own connections [55].<br />

– Mining and Exploiting Structure – social networks also can be formed by applying transformations<br />

on other, typically bipartite, graph representations identified in datasets.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!