|Thesis abstract: |
Even though search systems are very efficient in retrieving world-wide information, they can not capture some peculiar aspects and features of user needs, such as subjective opinions and recommendations, or information that require local or domain specific expertise. In these scenarios the knowledge of an expert or a piece of advice from a friend can be more useful than any factual information retrieved by a search system. This way of exploiting human knowledge for information seeking and computational task is called Crowdsourcing. People may take part to Crowdsourcing for a variety of motivations, which include non-monetary ones, such as public recognition, fun, or the genuine wish of contributing their knowledge to a social process. Various research works applied this paradigm to different fields and within various communities, including information retrieval, databases, artificial intelligence and social sciences.
The main objective of this work is to develop methodologies for the creation of application based on Crowdsourcing and social interaction.
The final outcome will be a framework based on model-driven approach that will allow end user to develop their own application with a fraction of the effort required by the traditional approaches.
This framework will guarantee a strong control of the execution of a crowdsourcing task by mean of a declarative specification of objectives and quality measures.
In order to achieve the objectives a Model Driven approach will be used, mainly for two reasons: first it allows an higher level of abstraction being platform agnostic, second it allows quick protoyping and multiple perspective design of the crowdsourcing campaign, thanks to several models and model transformations that reduce the development effort and lead to automatic code generation.
The ultimate goal is to automatically select the optimal strategies and generate the control rules given the domain model and the problem model. To achieve this, specific metrics are needed in order to evaluate the effectiveness of a selected approach. Thus part of this work will address the research and definition of empirical way to evaluate the optimality of a specific strategy. In parallel to this activities a concrete prototype will be developed. It will
be a web application that will allow the creation and execution of task on various platforms. It will allow the definition of strategies and control rules. Finally it will implement evaluation metrics in order to automatically select the optimal strategies given the problem.
Validation of the approach will consist of quantitative and qualitative analysis of results and performance of the system upon some sample scenarios, where real users from social networks and crowdsourcing platforms will be involved.