Web 2.0 applications emergence and increasing popularity, has influenced users roles which have become more dynamic and interactive, since users on the Web now act as co-developers and content regulators. Social interactions produce large scale of evolving information which can be exploited towards improving several popular applications (recommendations, trends detection etc). This social dimension was emphasized in the Web 2.0 or Social Web technologies and applications resulting in a remarkable bursting of web usage and content availability, and highlighting, at the same time, the need for efficient techniques’ implementation towards exploiting this collective knowledge.
The Intelligent Web 2.0 Recommendation Module may be adjusted to different web 2.0 applications, and efficiently extract users profiles by applying advanced mining techniques and considering various features such as users’ topics of interest, time locality, positive or negative reactions to social events etc.
The Intelligent Web 2.0 Recommendation Module bases its operation on a configuration file which makes it portable and adjustable to different web 2.0 applications. It consists of four components which are responsible for the following operations:
- Data Collection and Preprocessing ata fields described in the configuration file constitute the basis of information collected from the web 2.0 application and stored in suitable format. Then, efficient preprocessing operations are applied in order to construct convenient data representations in the predefined feature space (described in the configuration file) which will next be used in the mining process.
- Data Analysis:Clustering and community detection techniques will be applied on the collected data in order to identify users’ interests and extract their profiles. These techniques will consider the semantic and social aspects of users’ data as well as time, geographical information and means of access. Relations between users will be estimated using metrics that contemplate the former information.
- Evaluation of the extracted communities:information analysis methodologies and practices result in communities whose characteristics contribute to:
- Users’ profile extraction
- Communities identification
- Trends detection
- Recommendations: based on the extracted communities, an efficient and accurate recommendation mechanism can be supported and in particular, the recommendations may concern:
- Users: propose information relative to their interests.
- System Administrators: provide more personalized services.
The tool which will be created will integrate the operation of the four previously described components providing an autonomous and easy to use module for efficient recommendations.