Periodic Report Summary 2 - DIP3 (The 3Ps of Distributed Information Delivery: Preferences, Privacy and Performance)
Today, there is an abundance of data on line. The grand challenge is turning this huge amount of data to knowledge useful to the individual users of the Internet. DIP3 addresses this challenge by tackling one form of data processing, often referred to as 'push' data delivery. In push data delivery, instead of explicitly searching for information, users get notified when relevant information becomes available. Examples of such systems include RSS feeds, news alerts and aggregators. The scientific objective of the DIP3 project is to derive models, algorithms and techniques to control both the amount and quality of information received by users.
To this end, DIP3 introduces user preferences in data delivery to rank data items based on their relevance to the users. Although preference specification has been extensively studied, there is little previous research work on incorporating preferences in Internet-scale data delivery. Furthermore, DIP3 exploits the inherent social connections between users in Web 2.0 as expressed through social networks, social tagging, and other community-based features to enhance preference specification and ranked information delivery. Summarising DIP3 facilitates delivering to the users the most relevant and interesting information. To achieve its objectives DIP3 has produced research results of high quality and subsequently published them in top international conferences and journals.
Whereas traditional pub / sub systems rely on a binary, match / no-match model for sending relevant data to users DIP3 proposes non binary matching, where items are assigned degrees of relevance. Furthermore, appropriate data structures were proposed for distributed redundancy elimination. To address the abundance of online-data, DIP3 has proposed novel ways to explore databases through recommendations. Database recommendations extend the results of database queries with additional related information highly correlated with these results. The representation, composition and application of preferences in databases were advanced through new data structures that result in improving performance of managing database preferences.
Research has also focused on preferences in conjunction with keyword-based search in relational databases. To increase the quality of the search results, two new metrics were introduced that evaluate the goodness of the result as a set, namely coverage of many user interests and content diversity. New research results were attained in the context of database selection for XML document collections. The focus of this research is on keyword queries with Lowest common ancestor (LCA) semantics for defining query results. To improve performance, both in terms of storage and processing efficiency, appropriate summaries of the LCA information based on bloom filters were proposed.
Finally, although there is a lot of research on privacy, privacy-preserving push-based delivery has not been explicitly addressed by previous research. DIP3 has derived a set of new personalized privacy models and mechanisms. Knowledge transfer to university level through offering a new graduate course on data privacy in social networks, adapting two undergraduate courses on databases and data mining and supervising three MSc students, one PhD student and one undergraduate student. Knowledge transfer to European level through new research collaboration with University of Cyprus and University of Pittsburgh (joint supervision of graduate and PhD students), IBM (joint work on 'privacy in publish/subscribe systems'), University of Bari and University of Patras (joint project under the territorial cooperation program).
The research results of DIP3 facilitated the launch of: two new projects aimed at enhancing regional SMEs through advanced distributed information delivery tools through social networks (INTERSOCIAL) and smart phones (Epirus on Androids).
To this end, DIP3 introduces user preferences in data delivery to rank data items based on their relevance to the users. Although preference specification has been extensively studied, there is little previous research work on incorporating preferences in Internet-scale data delivery. Furthermore, DIP3 exploits the inherent social connections between users in Web 2.0 as expressed through social networks, social tagging, and other community-based features to enhance preference specification and ranked information delivery. Summarising DIP3 facilitates delivering to the users the most relevant and interesting information. To achieve its objectives DIP3 has produced research results of high quality and subsequently published them in top international conferences and journals.
Whereas traditional pub / sub systems rely on a binary, match / no-match model for sending relevant data to users DIP3 proposes non binary matching, where items are assigned degrees of relevance. Furthermore, appropriate data structures were proposed for distributed redundancy elimination. To address the abundance of online-data, DIP3 has proposed novel ways to explore databases through recommendations. Database recommendations extend the results of database queries with additional related information highly correlated with these results. The representation, composition and application of preferences in databases were advanced through new data structures that result in improving performance of managing database preferences.
Research has also focused on preferences in conjunction with keyword-based search in relational databases. To increase the quality of the search results, two new metrics were introduced that evaluate the goodness of the result as a set, namely coverage of many user interests and content diversity. New research results were attained in the context of database selection for XML document collections. The focus of this research is on keyword queries with Lowest common ancestor (LCA) semantics for defining query results. To improve performance, both in terms of storage and processing efficiency, appropriate summaries of the LCA information based on bloom filters were proposed.
Finally, although there is a lot of research on privacy, privacy-preserving push-based delivery has not been explicitly addressed by previous research. DIP3 has derived a set of new personalized privacy models and mechanisms. Knowledge transfer to university level through offering a new graduate course on data privacy in social networks, adapting two undergraduate courses on databases and data mining and supervising three MSc students, one PhD student and one undergraduate student. Knowledge transfer to European level through new research collaboration with University of Cyprus and University of Pittsburgh (joint supervision of graduate and PhD students), IBM (joint work on 'privacy in publish/subscribe systems'), University of Bari and University of Patras (joint project under the territorial cooperation program).
The research results of DIP3 facilitated the launch of: two new projects aimed at enhancing regional SMEs through advanced distributed information delivery tools through social networks (INTERSOCIAL) and smart phones (Epirus on Androids).