Random walks on the reputation graph

• Autor(a) principal: Sabir Ribas
• Data de Publicação: 2017
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Repositório Institucional da UFMG
• Texto Completo: http://hdl.handle.net/1843/31006
• topic: Random walks on the reputation graph

The identification of reputable entities is an important task in business, education, and in many other fields. In general, the reputation of an entity reflects its public perception, which touches upon a variety of aspects that may impact the identity of the entity, such as its prowess, integrity, and trustworthiness. Indeed, more reputable entities are presumably a better fit for most purposes. Thus, while reputation is a widespread notion in society, it is albeit an arguably ill-defined one. As a consequence, quantifyingreputationischallenging. Indeed, existingattemptstoquantifyreputation rely on either manual assessments or on a restrictive definition of reputation. Inthisthesis,insteadofrelyingonasingleandprecisedefinitionofreputation,we proposetoexploitthetransference ofreputationamongentitiesinordertoidentifythe most reputable ones. To this end, we introduce a conceptual framework of reputation flowsandproposeametricbasedonit, whichwecallP-score. Thisframeworkconsists of a random walk model that allows inferring the reputation of a target set of entities with respect to suitable sources of reputation. By using it, we can better understand how reputation flows between distinct entities in a reputation graph. Weinstantiateourmodelinanacademicsearchsettingtoaddressthreecommon ranking tasks namely, research group ranking, author ranking, and publication venue ranking. By relying on publishing behavior as a reputation signal, we demonstrate the effectiveness of our model in contrast to standard citation-based approaches for identifying reputable venues, authors, and research groups in the broad area of Computer Science. In addition, we demonstrate the robustness of our model to perturbations in the selection of reputation sources. Finally, we show that effective reputation sources can be chosen via the proposed model itself in a fully automatic fashion.