The Web has connected people in ways that were unimaginable just a few years ago. In particular the remarkable rise of social networks has increased the possibility of staying in contact with family, friends, colleagues and even bare acquaintances. It has created an environment for collaborative working. Real-time Web has allowed us to experience the presence of others while on-line and share our lives with them. Most importantly, the Web has integrated itself in our everyday activities. We communicate, we buy, we publish, we read, we give presentations, we organize content, we rate, we mate … on the Web.
hypios, a young marketplace for solutions, wants to push the boundaries on the ways the Web can help us innovate. We want to bring ideas and solutions–one’s whose full extension and applicability had not yet been discovered–out of their obscure hiding places and help them shine in the light of their utility. This is a story about our progress on this mission and the use of Semantic Web technologies to help get us there.
Hypios is a marketplace for solutions. We’re one of the first problem-solving Social Networks on today’s Web. Companies facing R&D problems use hypios to externalize their problems and broadcast them to a larger research audience, which then finds innovative and unexpected solutions to these problems. R&D departments usually have expertise in a certain domain, and approach problems from a certain perspective. But it’s logical to expect that somewhere on the Web there are people with different perspectives who can approach the problems differently and come up with something unexpected. The goal of problem marketplaces is to ensure that R&D problems reach the right people on the Web.
In his or her Web-life, a typical user writes blog posts, tweets, creates slideshare presentations, writes blog comments, makes friends, organizes content using tags, reads RSS feeds, and so on. All those activities leave valuable traces about his interests and his knowledge. We can divide these traces into five categories:
Content Created by User: blog entries, Tweets, research publications, comments, mailing list and forum posts
Content About the User: pages about the user, accredited sources confirming her experience and education, other pages that build her reputation
User’s Activities: participation in projects, taking roles in scientific and professional events
Social Graph: Who does the user know? Who follows the user?
User Profile and CV Data: basic interests, experience, affiliations
Social Semantic Web research efforts have already provided vocabularies for expressing many of those types of data. Lightweight ontologies like FOAF, DOAP, DOAC, SIOC, and OPO allow us to publish this information in a semantically rich, interlinked form, giving way to data integration, inference, and interoperability – the great promisses of the Semantic Web.
If we look at the Linked Open Data (LOD) Cloud, we can discover that a lot of useful data from those categories has already been made available in RDF form. In order to bring some structure to the uniform representation of the LOD Cloud, we have created a map of existing LOD data sources by categories of expertise-related data they contain.
Map of existing LOD data sources by categories of expertise-related data
See a larger image HERE
The map shows significant coverage of some types of information, but also identifies empty spaces where it is possible to imagine tools for transforming existing Social Web data to semantic forms. For instance public mailing lists and Question and Answers sites are a good source of data for expert-identification, especially in those cases where posts are rated by users. Making data from these sources available in semantic form would facilitate the creation of general Linked Data-based expert-finding approaches. The same sitation holds for Academic and Professional events where, currently, real semantic data exists only for Semantic Web conferences (via Semantic Web Dog Food website), but not for other fields. On the other had the existing data shows how easiliy one can identify competent people by querying for authors of publications in the domain, or for chairs of program committees. Semantic Web Dog Food is actually a good prof of concept for expert search that sets the example for other domains. A more detailed analysis of exploitability of existing LOD soruces for expert search and examples of possible queries giving the lists of experts is presented in our paper “Looking for Experts? What can Linked Data do for You?” on the Linked Data on the Web 2010 workshop.
As opposed to earlier expert search approaches that worked with legacy data, the Linked Data approaches benefit from increased flexibility. The data in the Linked Data is to be provided in a perspective-agnostic form, i.e. with many possible uses and perspectives in mind. On the contrary, legacy data approaches used to define a certain expertise metric (e.g. experts are people who wrote research articles on a certain topic) and then extract and formalise only the data needed for this specific metric. For instance an expert-finding approach might extract and use data from research publications (their topics and authors) but initially omit the conference and date information. Further approaches would then be unable to reuse the same data for metrics taking into account the the conferences where the authors we present and the dates when they started to publish articles in a certain domain. Although simplistic, this examples shows a real phenomenon that we have discovered in our analysis of expert search approaches. Changing this approach after the fact proves difficult. With the Linked Data approach, since data is created for multiple purposes we can easily manipulate the data-views, swiching from looking at authors of research publications to the members of program committees at various research events. When data is published in Linked Data form it is generally perspective-agnostic, and supports various views on the same entity (e.g. a person) and varios scenarios of use.
Linekd Data also gives us access to hubs of general-purpose knowledge like DBPedia and Freebase, which facilitate work with expertise domains by helping us define categories in standard ways. Also, these sources provide information on the “closeness” of topics (e.g. their family resemblance), and thus allow us to search for experts in closely resembling domains or domains that are broader than the domain in which the problem was initially found.
Semantic Web standards also make it easy to identify the equality of expertise traces (like research papers, blogs, etc.) on various sources.
At hypios we explore how can me make use of all those benefits to bring R&D problems closer to their potential solvers. We seek to find domain-experts, people interested in topics related to the problems; all using the traces available as Linked Data on the Web. Our efforts will soon expand in filling the gaps that exist in the current LOD cloud to make expert finding even better supported.