Authors:
Stanley Loh
1
;
Fabiana Lorenzi
2
;
Roger Granada
3
;
Daniel Lichtnow
4
;
Leandro Krug Wives
5
and
José Palazzo Moreira de Oliveira
5
Affiliations:
1
UCPEL Universidade Católica de Pelotas;ULBRA Universidade Luterana do Brasil, Brazil
;
2
ULBRA Universidade Luterana do Brasil;UFRGS Universidade Federal do Rio Grande do Sul,, Brazil
;
3
UCPEL Universidade Católica de Pelotas, Brazil
;
4
UCPEL Universidade Católica de Pelotas;UFRGS Universidade Federal do Rio Grande do Sul, Brazil
;
5
UFRGS Universidade Federal do Rio Grande do Sul, Brazil
Keyword(s):
User profile, User profile similarity, Collaborative recommender systems.
Related
Ontology
Subjects/Areas/Topics:
Data Engineering
;
Ontologies and the Semantic Web
;
User Modeling
;
Web Information Systems and Technologies
;
Web Interfaces and Applications
;
Web Personalization
Abstract:
This paper presents investigations on representing user’s profiles with information extracted from their scientific publications. The work assumes that scientific papers written by users can be used to represent user’s interest or expertise and that these representations can be used to find similar users. The goal is to support similarity evaluations between users in a model-based collaborative recommender. Representing users by their publications can help minimizing the new user problem. The idea is to avoid the necessity of asking users to evaluate a set of items or give some information about their preferences, for example. In scientific communities, particularly on digital libraries and systems focused on the retrieval of scientific papers, this is an interesting feature. We have conducted some experiments to compare different techniques to represent the papers (title, keywords, abstract and complete text) and two kinds of text indexes: terms and concepts. Furthermore, two distin
ct similarity functions (Jaccard and a Fuzzy function) were applied on these representations and then compared with the goal of finding similar users.
(More)