While the first generation of e-participation has been based on official e- participation spaces owned and operated by government, the second one is oriented towards exploiting the highly popular web 2.0 social media for performing `crowdsourcing' of policy-related knowledge, opinions and ideas from citizens, through posting relevant policy-related content to some social media and then retrieving and processing citizens' interactions with it. Recently, the idea of a third generation of e-participation has been proposed, which is based on a more `passive' form of crowd-sourcing in social media, through automated passive search by government agencies for content on a public policy under discussion, that has been created in a large number of predefined relevant web 2.0 sources (e.g. political blogs, news websites, facebookand twitter accounts) by citizens freely, without any direct stimulation by government, retrieval and sophisticated processing of this content. In this paper we analyze and elaborate this idea, based on cooperation with potential users experienced in the design of public policies, through a combination of quantitative and qualitative techniques. Initially, the main roles required for the practical application of this concept are identified, and then the functional requirements of each of them are determined. Finally, based on these functional requirements the architecture of a central platform supporting the application of this concept is designed.