| Abstract |
Social networks are constantly evolving to support the increasing needs for knowledge
sharing, interaction and collaboration among people from all over the world through
Web. These networks provide to users many interaction platforms where they can share
their opinions and their life experiences. Debate portals are one type of such platforms
where people can express their views in the form of arguments and participate in support
of or against issues occurring in the dialogue in a more structured way. As long as social
networks and other sources of discussion produce more and more content, the need to
understand and summarize the opinions expressed within dialogues increases, in order
to reduce the burden of having to go through the entire debate.
The system provides a debating environment that aims to motivate people to participate
in structured, goal-oriented dialogues. As a debating platform, it enables users to raise
issues, ask their own questions, post supporting or counter-arguments, comment and
vote. The overall objective is to offer different means of analysis of the debates, in order
for the participants to obtain a complete picture of the validity and justification strength
of each individual opinion expressed, as well as of the acceptance of the positions issued
within each debate. The system provides a range of functionalities, the most important
of which concern the creation of new topics of discussion, the evaluation of arguments
with different metrics, and the analysis of various aspects of the dialogues.
In this work, we start with providing an argument map for modeling discussions and the
relations between them with debate elements such as issue, position, pro-argument and
con-argument. Then, we apply an existing formal framework for evaluating the strength
of arguments, called sm-Dice. Every argument strength is calculated based on a multi-aspect evaluation. Next, we implement a debate analysis by taking into account the
various aspects of the dialogues. This analysis covers different information needs
emerging from users, in order to summarize various aspects of a debate, focusing not only
on arguments, but also on user profile characteristics throughout the decision-making
process. A collection of machine learning algorithms is applied for the clustering of
features and the extraction of association rules, such as the Kmeans and Apriori
algorithms.
Our system uses an RDF ontology for representing the argument map of any
dialogue,
stored as RDF-triples in the Virtuoso repository. The user interface is designed with Web
technologies, whereas the Server Tier is implemented with Servlets and Java API classes.
|
Search
