resulting relevance, thus helping to decrease the division between the potential endorsement that an opinion could have and its effective endorsement (see Figure 3). A wide variety of research needs to be carried out also in this context.
Figure 3 - Opinion Refinement Relevance
Both the discrete rounds procedure (with Pareto front) and the continuous time (with sampling) procedure have their positive side and their shortcomings. We shall investigate and implement both of them, analyse their differences, and then understand when is it better to use one or the other.
In every moment, users should be allowed to investigate the phylogenetic tree of opinions that they (or others) produced by a particular opinion, find out in each step why was an opinion favoured respect to another and how many people endorsed each. This data should be given both as raw data in a particular form (e.g. XML) and in a suitable graphical form. Raw data should permit any user to download, and reprocess the information, thus adding an extra layer of transparency.
Knowledge Learning. As opmions are collected, EUGAGER users will be allowed to search for them and archive them by adding to them new keywords (called tags). Thus each user will potentially be allowed to add her keywords to each opinion (even not hers). This is the base of what is called a broad folksonomy. Users will then be allowed to retrieve the opinions they have stored under a certain tag (or sets of tags). Using the (weighted) emergent set of tags, for each opinion we shall apply a measure that given every two opinions will find their nearest neighbours, thus letting users move from opinion to opinion.
We also shall implement a search box. To this aim in the following we explain how we should organize information. Users will be allowed to insert a set of keywords. Those words will be evaluated as a weighted set, their weights following a power law distribution and with steepness equivalent to the steepness of tags in a broad folksonomy. Using this information, the most relevant results will be returned. Since nearest neighbours opinions are linked to each other's, we can represent the set of opinions as a graph. This graph will not necessarily be connected. But each connected component (called cluster) will represent a set of opinions that can be reached moving through nearest neighbours, which could be a search option. Finally, we will present the set of clusters as a form of top down search. Note that this set of clusters will allow the user to look for broad areas of opinions. Moreover, clusters will give some feedback on the culture of the users having generated them, on which terms constitutes their ontologies and on which terms are considered related to each other.
To achieve the objective described in 1.1.2, we will main focus on the following properties that will be considered as requirements for the design of the EUGAGER platform.
• Authentication and Identity Management. The provision of assurance of the claimed identity of an entity.
• Availability. The property of being accessible and usable upon demand.
• Confidentiality. The property that information is not made available or disclosed to unauthorized individuals, entities, or processes.
• Integrity. The property that information have not been altered or
destroyed in an unauthorized manner.
A detailed threat analysis will be performed in the initial phase of the project, to identify the inherent risks of our architecture; through the identification of the possible threats and vulnerability of the architecture, we will identify the possible attacks to the system (such as,