Modeling Scholar Profile in Expert Recommendation based on Multi-Layered Bibliographic Graph

Purwitasari, Diana (2020) Modeling Scholar Profile in Expert Recommendation based on Multi-Layered Bibliographic Graph. Doctoral thesis, Institut Teknologi Sepuluh Nopember.

Text 07111660012001-Dissertation.pdf Download (6MB)

Abstract

A recommendation system requires the profile of researchers which called here as Scholar Profile for suggestions based on expertise. This dissertation contributes on modeling unbiased scholar profile for more objective expertise evidence that consider interest changes and less focused on citations. Interest changes lead to diverse topics and make the expertise levels on topics differ. Scholar profile is expected to capture expertise in terms of productivity aspect which often signified from the volume of publications and citations. We include researcher behavior in publishing articles to avoid misleading citation. Therefore, the expertise levels of researchers on topics is influenced by interest evolution, productivity, dynamicity, and behavior extracted from bibliographic data of published scholarly articles. As this dissertation output, the scholar profile model employed within a recommendation system for recommending productive researchers who provide academic guidance. The scholar profile is generated from multi layers of bibliographic data, such as layers of author, topic, and relations between those layers to represent academic social network. There is no predefined information of topics in a cold-start situation, such that procedures of topic mapping are necessary. Then, features of productivity, dynamicity and behavior of researchers within those layers are taken from some observed years to accommodate the behavior aspect. We experimented with AMiner dataset often used in the following bibliographic data related studies to empirically investigate: (a) topic mapping strategies to obtain interest of researchers, (b) feature extraction model for productivity, dynamicity, and behavior aspects based on the mapped topics, and (c) expertise rank that considers interest changes and less focused on citations from the scholar profile. Ensuring the validity results, our experiments worked on standard expert list of AMiner researchers. We selected Natural Language Processing and Information Extraction (NLP-IE) domains because of their familiarity and interrelated context to make it easier for introducing cases of interest changes. Using the mapped topics, we also made minor contributions on transformation procedures for visualizing researchers on maps of Scopus subjects and investigating the possibilities of conflict of interest.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	modeling scholar profile, bibliographic data for academic social network, productivity and dynamicity features, behavior-based features, expertise rank
Subjects:	Q Science > QA Mathematics > QA166 Graph theory Q Science > QA Mathematics > QA274.7 Markov processes--Mathematical models. Q Science > QA Mathematics > QA76.76.E95 Expert systems T Technology > T Technology (General) > T58.62 Decision support systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20001-(S3) PhD Thesis
Depositing User:	Diana Purwitasari
Date Deposited:	08 Sep 2023 01:21
Last Modified:	08 Sep 2023 01:21
URI:	http://repository.its.ac.id/id/eprint/75206

Actions (login required)

View Item