How can greater exposure of social data in semantic web formats be a game-changer? Peter Mika, researcher, search at Yahoo!, has a lot of ideas on the subject. Mika, the author of Social Networks and the Semantic Web, will be speaking at a panel at next week’s Web 3.0 conference on the Semantic Web for Social Media, and he took some time to explore some of the issues around the topic with us. Read on:

SemanticWeb Blog: Is more social data being exposed on the web in semantic web formats, and how can this be leveraged/analyzed to create new value for businesses/other organizations/consumers?

Mika: Most social data is unfortunately still either walled off completely, available in HTML only or exposed through proprietary APIs returning non-semantic formats.

There has been some consolidation both in technology and market terms, which makes the integration of particular services easier and more efficient. For example, technological "glues" such as the Yahoo Query Language are enabling the average developer to integrate four or five of his/her favorite APIs within an hour. However, providing services that integrate information on a global scale would still run into the problems of insufficient data and incompatible APIs. As an example, major search engines still cannot perform the kind of high precision people search that semantic technologies should enable. (A search for Peter Mika still intermixes results related to the computer scientist and the ballet dancer.)

Still, there are opportunities, and they are everywhere simply because "social" is not a business or a product, it's a dimension of many human activities. Taking shopping/services as an example: It is clear that when searching for a restaurant, I would give a preference to those places visited and liked by my friends (or where they are eating at the moment!). Another example is news. Twitter is an excellent showcase of how much news is a social construct. Just like in real life, where our opinion on what events of the world are important are shaped by the real friends we talk to, our Twitter friends influence us in our online news consumption.

The opportunity is that by looking at online data, where we know the networks of users, and their identities are correlated with streams of activity related to particular objects or types of objects, we can precisely study how these influences work. What kind of friends influence us? In what way? In what situations? Regarding what types of objects? How do these influences develop? Once we understand how it works, we should be able to teach computers to be "socially intelligent", without necessarily building full artificial intelligence.

Fragmentation is the largest problem in my view. Not only in terms of explicit relationships to other users, but more critically in terms of interests and activities that form the basis for relationships. I'm saying that this latter is more critical in that I might call officially all my Facebook friends as friends, but the relationship history (the messages sent and received, the photos, the
likes/dislikes) would give a much more refined picture of the relationship graph. Currently, not just my explicit social graph is spread out over the Web, but all the little bits of information that constitute social activity.

SemanticWeb Blog: Is the vocabulary problem also a challenge in leveraging these opportunities?
Mika: The vocabulary problem on the Semantic Web is still unsolved in general, but with more usage it takes less time for "clear favorites" to emerge.
In the social space, the vocabulary problem is now reduced to choosing between a small number of formats. Nevertheless, and this has been discussed at a number of different forums, the Semantic Web would really benefit from a more centralized approach to vocabulary management.

SemanticWeb Blog: How do you see challenges being addressed today, broadly (by industry efforts such as the W3C Social Web Incubator Group) and by Yahoo specifically?

Mik: As far as I have managed to follow their activity, the W3C Social Web Incubator group was the starting point for many interesting discussions (their meeting notes are available online), because of their method of inviting an external speaker to each of their meetings. I've also attended one of the meetings as an invitee, and brought to the group's attention the existence of two incompatible RDF ontologies for VCard.

The problem again from a search engine perspective is that they are different models of the original vCard specifications, and data sets using these different ontologies are not compatible without fairly complex mappings. These proposals were then merged, and one of the ontologies was deprecated. My colleague Eran Hammer-Lahav also attended one of the meetings, and talked about the problem of discovery, i.e. the various existing and rather different mechanisms to find the metadata related to a web resource .

Yahoo is on the forefront of implementing open web standards for interoperability (OpenID, OAuth), and has been also the driving force behind the Open Web Foundation.

SemanticWeb Blog: What's the most exciting thing you've personally been involved with at Yahoo around semantic search in the last year?

Mika: It's actually a lesser known feature of the new Yahoo Search result page which shows up for particular queries: The possibility to drill down based on the type of resources returned. Try "ipod" and you will see the option "Shopping Sites" on the left. This restricts the selection of results to those that we know contain product information because of explicit markup (e.g. GoodRelations markup) or through information extraction.

SemanticWeb Blog: Looking at the coming year, what are your predictions about how the synergies between social networks and the Semantic Web will evolve? What is your greatest hope/biggest concern?

I think we will see more services that will exhibit "social intelligence" as described above, particularly in the mobile area. Unlike desktops, mobile phones are personal and trusted devices. Also, when using the mobile, our information needs are much closer to what currently the semantic web provides, i.e. precise factual information about the objects surrounding us, put into the context of other objects and the information on the Web.

I find very encouraging the increasing business interest in the Semantic Web area, especially from the side of online media and content publishing. My concern is a somewhat growing distance between the more business oriented "Semantic Technology" sector and the academic Semantic Web community, which means that practical needs have less and less influence on the research agenda, and research ideas have less of a chance to find realistic applications.