Managing Structured Collections of Community Data

Data management is becoming increasingly social. We observe a new form of information in such collaborative scenarios, where users contribute and reuse information, which resides neither in the base data nor in the schema information. This “superimposed structure” derives partly from interaction within the community, and partly from the recombination of existing data. We argue that this triad of data, schema, and higher-order structure requires new data abstractions that – at the same time – must efficiently scale to very large community databases. In addition, data generated by the community exposes four characteristics that make scalability especially difficult: (i) inconsistency, as different users or applications have or require partially overlapping and contradicting views; (ii) non-monotonicity, as new information may be able to revoke previous information already built upon; (iii) uncertainty, as both user intent and rankings are generally uncertain; and (iv) provenance, as content contributors want to track their data, and “content re-users” evaluate their trust. We show promising scalable solutions to two of these problems, and illustrate the general data management challenges with a seemingly simple example from community e-learning (“ce-learning”). 1. A VISION: MASSIVE COMMUNITY E-LEARNING WITH PAIRSPACE We will argue that management of collections of community data requires a new abstraction that does not fit well in the common dichotomy of data and schema information. We illustrate this idea with the vision of a massive online question-answer learning community of users, grouped around a hypothetical tool we refer to as PairSpace. We prefer to keep the overall setup simple. This is a concrete community data management scenario that illustrates the main issues in this paper, while at the same time, seems to have a simple relational implementation. Note that the underlying challenges naturally extend to more complex and general community content management scenarios. 5 Biennial Conference on Innovative Data Systems Research (CIDR ’11) January 9-12, 2011, Asilomar, California, USA. This article is published under a Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0/), which permits distribution and reproduction in any medium as well allowing derivative works, provided that you attribute the original work to the author(s) and CIDR 2011. 1 day 3 days 1 week 1 month 6 months correct