Article Preview
Top1. Introduction
Web communities are nowadays one of the most used sources of finding experts by organizations. Social networks, with the huge volume of data generated by users are considered as a fruitful field to identify competencies and experts. The data is available under the form of posts, threads, tags and user information is transferable into semantic form (De Vocht et al., 2017). Searching the human expertise has recently attracted considerable attention. Since the human expertise is hard to formalize (Cifariello et al., 2019) and the absence of generic expertise formalizing rules (Petkova & Croft, 2008), expert finding is perceived by researchers as a computationally challenging task. It has been an active research area in many fields such as artificial intelligence, knowledge management, computer-supported cooperative work, and others (Al-Taie et al., 2018).
Expertise retrieval (or search) consists of two main tasks expert finding and expert profiling. While an expert finding task answer the question (“Who are the experts on topic X?”), expert profiling task is interested on (“What topics does person Y know about?”). Expert finding received considerably more attention than the expert profiling (Rybak et al., 2014). The main purpose of expert finding is indeed, to determine and rank persons likely to be expert in a specific area, whereas the expert profiling objective is to determine skill areas in which a person is specialized (Dehghan et al., 2019a). Expert finding systems use implicitly or explicitly provided data about user’s expertise to identify appropriate experts. However, creating user profiles on their personal content and feedbacks is a challenge. In addition, self-declarations of skills can be incorrect, inaccurate, or insufficient. In other words, people may not be aware of having a certain skill at a level of proficiency or they may lie on their descriptions of what they contributed or accomplished (Fazel-Zarandi & Fox, 2011). An alternative task, built on the same underlying principle of computing people–topic associations, is expert profiling, in which systems have to return a list of topics that a person is knowledgeable about (Balog & De Rijke, 2007; Berendsen et al., 2013a).
Valuable sources of information are usually used to identify the user’s expertise level. The expertise related data sources contain a heterogenous documents such as candidate’s publications, technical reports, emails and web pages. It is extracted from two sources, social networks (Wang et al., 2013) and documents (Balog et al., 2006), so both content and communication links are exploited as sources of evidence or indicators.
The social tagging activities have been integrated in several works as information related to candidate expertise (Budura et al., 2009; Serdyukov et al., 2011). Social tagging has emerged in the social web for quite a few years, as a support to the organization of shared resources by allowing users to categorize and find these resources. Social data can also be a support for enhancing the personalized search (Zhou et al., 2016) by modeling user profile from social tagging. The term tagging which is usually associated with folksonomy, refers to a classification (taxonomy) made by users (Folks) (Marlow et al., 2006; Mathes, 2004). It arise from data about how people associate terms with content that they generate, share, or consume (Gruber, 2007). It is also defined by (Broudoux, 2006) as a series of metadata created by a collective users to categorize and retrieve online resources. Social Tagging is recognized for its potential to leverage collaborative production of information that support a wide range of mechanisms such as social search (Brusilovsky et al., 2018; Xie et al., 2016), profile extraction (Kasper et al., 2017; Kichou et al., 2016) and recommendation (BellogíN et al., 2013; Font et al., 2013; Kichou et al., 2016).