As discussed hitherto, there is a great body of literature which encompasses various approaches in bridging the gap between folksonomies and ontologies in one (from social to formal) or the inverse (from formal to social) direction.
To the best of our knowledge there are few works that combined both direc-tions together: also in these cases, the role of users has not been evaluated deeply, remaining mostly underestimated.
Stemming from the ontology maturing process, the research of Van Damme et al. [DHS07] aimed to propose an innovative way that combine the strengths of col-laborative approach to ontology engineering with a mash-up of available social data and semantic resources. This research underlined the need for a semi-automated approach to construct a FolksOntology in which, firstly, folksonomies and their associated data, online lexical resources and ontologies and other Semantic Web re-sources, could be fully exploited for making ontologies. Secondly, all the information extracted from such resources should be validated by a community of users, stim-ulating the contribution of a collective human intelligence in order to enhance and improve results from previous stages. For instance, users could play a significant role in the discover of information and relations between tags not retrieved directly
from the resources. To ease users’ activities, the author proposed a set a possible functionalities that a FolksOntology should provide, such as proper visualization techniques as well as implicit and explicit voting mechanism on conceptual choices, in order to grasp the intentions of ontologies concepts. Unfortunately, the work did not present an evaluation framework aimed to validate the collaborative process, albeit it has been suggested to involve users continuously in the form of community approval of the ontology conceptualization.
The model proposed by Freddo et al. [FT09] concentrated primarily on how an ontology, previously generated from a folksonomy, could evolve. The main motiva-tion at the basis of this work was that folksonomies were generally not considered as useful information sources especially during ontology evolution process. This model could be distinguished into two subsequent stages: firstly, the ontology learning from folksonomies and, secondly, the ontology evolution from folksonomies. To demon-strate their approach, a simple case study was presented. Considering a folksonomy related to the tourist domain, the SCOT tag ontology based on the Newman’s model [NAR05] was populated. After that, the user built an initial ontology taking into account both relations among pairs of tags (e.g. as instance of, has and is a) and proper meta-properties. In the second phase of the proposal, new tags from the folksonomy were included and aligned to the concepts contained already by the ontology, according to the highest computed similarity value. This model could be considered at a very initial stage and provided interesting insights, but as observed by the authors, a deeper evaluation largely lacked.
The work of Sharif [Sha09] proposed a simple ontology-of-folksonomy model in order to describe how i) different elements could act in a dynamic space and ii) how implicit relations emerged from implicit complex networks within folksonomies. To this extent, two sub-models were defined, each describing two distinct processes.
The first sub-model consisted in an ontology-based tagging process, aimed at the knowledge acquisition and representation. It demonstrated how a user can enrich the tagging process through available ontologies. In an initial stage, new tags were suggested by the user to the ontology through both uncontrolled and controlled tag-ging: hereby, semantic repositories (e.g. Swoogle11, OntoSearch12, DAML Ontology Library 13, Protg Ontology Library 14) came into play with tag/concept recommen-dation. Then, thanks to an interactive display, user might confirm the candidate tag, or suggest a new one. In such a way user were empowered to extend the ontol-ogy with new concepts on demand. At the end an ontolontol-ogy evolution module was defined to map tags into the ontology. The second sub-model was defined as an ontology-based folksonomy query expansion model, whose objectives were mainly devoted to knowledge discovery, thanks to proper visualizations and user interactions. To this extent, user could browse an alphabetic list of tags, concepts,
11http://swoogle.umbc.edu/
12http://www.ontosearch.org/
13http://www.daml.org/ontologies/
14http://protege.stanford.edu/download/ontologies.html
actors or groups rather than search a query. From the one hand, ontologies would improve precision and recall, while from the other hand, users envisaged tag space enrichment with semantic relations by exploring online ontologies. Despite the cen-trality of the role of user was indirectly suggested, a framework for the evaluation of how users could benefit from these two sub-models aimed at checking their va-lidity, was not designed. Moreover, the definition of an enhanced visualization of the folksonomy, especially when it grown, was rather limited. Recently, Alves et al. [AS13] proposed an interesting and still ongoing research study, defining a so-called Folksonomized Ontology - FO from now on. This approach put forward the idea of a semantic “travel” in both directions: from folksonomies to ontologies and vice-versa. Basically, the FO allowed to i) enhance tag disambiguation, tag sug-gestion and semantic similarity driving rich semantic-based matching, categorization and tag suggestion, and to ii) support the review and enhancement of the ontology through contextual data. The initial stages of the proposal followed a trail of com-mon techniques, namely: the information extraction, the semantic enrichment, the tagset-ontology mapping and the ontology evolution. In particular, starting with the meta-data extraction from the folksonomy, tags were filtered and clean in a pre-process phase, in order to define proper tagsets, each of them representing a folksonomic concept. Within the enrichment stage, the latent semantics from the folksonomic tissue was extracted and fused with the ontology. The outcome of the mapping phase consisted in a set of mapping edges: each edge stores the degree of similarity between the tagset and the concept. At the end a fusing phase produced the FO, taking into account enriched relationships among concepts. Again, the user played a predominant role in the ontology evolution and maintenance: thanks to the interaction with a visual tool, the user could support the ontology in its re-view, analysing data and visualizing the cases in which the collaborative knowledge indicated that the ontology needed for a revision and/or an enhancement. A graph-ical tool came in support to the user, who could interact with a segment of the FO, analyzing concept and relations among them, and compared them with rela-tions captured from the folksonomy. Unfortunately, at the current stage, the tool has been intended only to suggest modification, with no support to any automatic change or ontology editing. At any rate, this research has advanced a quite complete process that overcome the gap between ontologies and folksonomies, considering the end user a key element in a continuum process of a folksontology improvement.
5.2.4 Discussion
To summarize the overview about the analyzed approaches presented in section 5.2, we identify a set of tasks where users play a crucial role within these systems. In particular, table 5.3 indicates, where applicable, the presence of the following tasks:
1. validation of automated mappings (Val): users’ intervention consists in the validation of a previous, generally automated, mapping of tags into concepts
of the ontology;
2. creation of new concepts from tags (Cre): the back-end user is able to create new concept and make proper associations;
3. explicit mapping (ExMap): users contribute with an explicit concept-to-tag mapping, assign a tag to a concept of an existing ontology;
4. ontology extension with tags (Ext): users enrich existing ontology due to their social annotations within a system;
5. ontology editing and evolution (EdEv): the user is able to edit and support the evolution of the ontology;
6. collaborative ontology improvement (Col): users edit actively the ontology in a collaborative environment.
Table 5.3: Comparison of the bridge-the-gap approaches between folksonomy and ontology considering users’ tasks.
1.Val 2.Cre 3.ExMap 4.Ext 5.EdEv 6.Col
Sotomayor, 2006 [Sot06] 3
Passant, 2007 [Pas07] 3
Limpsen et al., 2010 [LGB+10] 3
Cantador et al., 2011 [CKJ11] 3
Lezcano et al., 2012 [LGBS12] 3
Mika, 2005 [Mik05] 3
Gendarmi and Lanubile, 2006 [GL06] 3
Braun et al. 2007 [BSW+07] 3
Liu and Gruen,. 2008 [LG08] 3
Gaˇsevi´c et al., 2011 [GZT+11] 3
Van Damme and Siorpaes [VDS] 3 3 3
Freddo and Tacla 2009 [FT09] 3 3 3 3
Sharif, 2009 [Sha09] 3 3 3 3
Alves and Santanch´e, 2013 [AS13] 3 3 3 3
All these approaches limited the analysis and the evaluation of advantages offered both to domain experts, and to end-users, as long as they are utilizing a specific application.
5.3 Summary
In this chapter we have presented the issue of “bridging-the-gap” between folk-sonomies and ontologies, concentrating our attention on the role of the user, with a special focus on:
• which benefits a user could gain from their peculiar strengths and which weak-nesses have to be taken into account;
• to what extent a user is get involved in their creation, maintenance and evo-lution;
• which are the processes, steps and practices commonly carried out by these’ re-search studies and how the user has contributed to the resulting folksonomized ontology or ontologized folksonomy;
• which are the major advantages that a user could derive from an application that relies on hybrid and combined approaches.
As showed within our discussion, folksonomies and ontologies worked better to-gether, providing the user with a wide range of features and capabilities that would be difficult to achieve otherwise, without this synergy.
However, we can underline the existence of still critical open issues, for instance the lack in the definition of a general frameworkthat exploits both social and do-main’s experts knowledge, focused on both user needs and requirements. We can also point out that in the majority of the cases, there is a lack in visualization and authoring tools, capable to ease browsing and searching for tags and resources, rather than potentially similar users. Moreover, both user contextual navigation is scarcely supported.
Concentrating on the role of both tourists and domain-expert users, and taking into account all the observations and findings gathered so far, in the next and final chapter we:
• propose a novel bridge-the-gap approach, based upon zz-structures, that sup-ports the integration of social and semantic knowledge;
• extend the formal model of FOLKVIEW discussed in chapter 2;
• improve and extend the proposed model and the designed framework provided by TOGO case study, analyzed in chapter 4.