Sci7 is here reviewing and extracting interesting observations and comments from a recent article by Peter Mika titled: Flink: Semantic Web technology for the extraction and analysis of social networks. The piece was published in the Journal of Web Semantics.
Insight into the challenges involved in extracting meaningful information on relationships between individuals from material available online is presented with reference to a network of semantic web researchers complied by the author. Associations between researchers were automatically extracted from sources such Google Scholar, conference attendance and mailing lists. Many online mailing list archives obscure email addresses, or enable users to elect not to have messages archived - the researchers here had direct access to raw emails. Extracting information on connections between people from sources which are the natural products of the everyday work of those involved is presented as an alternative to a centralised database such as that run by Friendster. The evolved Orkut social networking service is also referred to.
Presenting information in a machine readable format is discussed. An individual may for example provide links on a personal home page to others, and these could be identified using the “rel” relationship attribute of the link to describe the nature of the association between the individuals. The FOAF (friend of a friend) project provides an alternative structured machine readable format for a user to provide such information. The advantage of these latter two methods is that users (rather than organisations such as Friendster) retain control over their own information.
Some of the problems encountered with extracting meaningful information from search engines was discussed, limiting searches to the area of interest using the search terms semantic web or ontology was one technique used to reduce junk data from entering the system, another problem discussed was with those individuals who used alternative spellings and shortened forms of their names.
Two interesting applications of relationship data are considered, one is as a method of narrowing a search. Much as it makes sense for search for a tradesman to be conducted within geographical constraints, perhaps there are types of information for which a search within a subset of the web defined by social, or other, relationships would be desirable. A scientist might search to determine if any collaborators or collaborators of collaborators have a particular piece of analytical equipment for example. The second application is in terms of judging the importance / credibility / trustworthiness of a information on the web, if it is a site that is authored by or trusted by members of a social network that may influence a user’s perception of that site.
Consideration is given to the privacy implications of data mining for social interaction data, the author suggests that the use of personal information for the construction of networks such as those described here is likely to become the subject of much debate in the future. There is no discussion though on how the barrier between private and public information may shift or alter in importance over time.
Those creating web sites, and online services should consider how they could be using the opportunity to interact with social networks, both by providing information to them, and using information derived from them. The opportunities for marketing provided by online social networks are immense, from trend spotting and data collection, through advertising to specific groups and creating a buzz for new products.
DOI link to original article: 10.1016/j.websem.2005.05.006
Tags: Semantic Web, Semantic, Flink, Google Scholar.