We present the MaGe system, which helps users and developers to build maps of the Web graph. Maps abstract and represent in a concise and machine-readable way regions of information on the Web. The Web is a large and interconnected information space (usually modeled as a graph) commonly accessed and explored via navigation enabled by browsers. To cope with the size of this huge (cyber)space, Web users need to track, record and specify conceptual regions on the Web (e.g., a set of Web pages; friends and their interests; a network of citations), for their own use, for exchanging, for further processing. Users often navigate large fragments of the Web, to discover and isolate very few resources of interest and struggle to keep connectivity information among them. The idea of a map of a Web region is essentially that of representing in a concise way information in the region in terms of connectivity among a set of distinguished resources (nodes).
The idea of a map on the Web is to represent in a concise and comprehensive way connectivity information between pairs of distinguished nodes. Given a conceptual region G of information on the Web, there can be several maps of G with different level of detail (i.e., nodes and edges to be included).
Formally, let = (V ; E ) be a Web region, where V and E are the set of nodes and edges respectively. Then: u ! v denotes an edge (u; v) 2 E. u v denotes a path from u to v in .
Let N V . Then, u N v if and only if there is a path from u to v in passing through intermediate nodes in N . not Let VM V be the set of distinguished nodes of the Web region i.e., those that we would like to represent. = (V ; E ), Definition 1 (Map) A map M = (VM ; EM ) of that VM V and each edge (x; y) 2 EM implies x = (V ; E ) is a graph such y in .
A basic (and highly used) example of map of the Web are bookmarks. In this
case, VM is the set of nodes highlighted or marked, and EM = ;, that is, there
is no connectivity recorded among them. An important idea is that of a good
map, i.e., a map which represents connectivity among the distinguished nodes
and avoids redundant edges [
Good maps have the nice properties (i) uniqueness and (ii) low complexity of computation. Indeed, given a region = (V ; E ) and a set of distinguished nodes VM V there exists a unique good map M = (VM ; EM ) that is computable in O(jVM j (jV n VM j + jE j)) by an adaptation of the BFS algorithm.
Maps are built on top of regions of information on the Web. To automate the
process of generating regions, MaGe uses the NautiLOD [
MaGe has been implemented in Java and is available for download4. It includes two main modules: the selection and the abstraction modules. The first 4 The MaGe website: http://mapsforweb.wordpress.com Specification of the Region
Map Creation one is responsible for the implementation of the NautiLOD language. In particular, given a seed URI and an expression, this module retrieves a Web region and a set of distinguished nodes. The second module, given the Web region and the set of distinguished nodes leverages the map framework to build maps. The decoupling between selection and abstraction enables to use the two functionalities also separately. MaGe is endowed with a GUI, which is shown in Fig. 2. It includes four main tabs. The first one (Fig. 2 (b)) is used to specify the region via a NautiLOD expression. The second and fourth display the region retrieved in RDF and the expression endpoints, respectively. The third tab (Fig. 2 (a)) deals with the creation of maps and their visualization. Both regions and maps can be saved in RDF allowing their storage, sharing, reuse and exchange. We now provide an example that we plan to show (along with others) in the demo. A video explaining how to use the tool is available at http://youtu.be/BsvAiX3n968. Maps of Influence. An influence network is a graph where nodes are persons and edges represent influence relations. We leverage information taken from dbpedia.org and the property dbpprop:influenced.
Example 3 Build a map of a region containing people that have influenced, or have been influenced by Stanley Kubrick (SK) up to distance 6. The distinguished nodes must be scientists.
The region can be specified via the following NautiLOD expression. Here, the URI of SK in DBpedia (dbpedia:Stanley_Kubrick) is used as seed node: dbpprop:influenced<1-6>[ASK {?p rdf:type dbpedia:Scientist.}] In the expression, the notation <1-6> is a shorthand for the concatenation of (up to) six steps of the predicate dbpprop:influenced, while the ASK query in the test [ ] is used to filter the distinguished nodes (i.e., scientists).
The availability of machine-processable information at a Web scale opens new perspectives toward the development of systems for the harnessing of knowledge on the Web. We contend that maps, key devices in helping human navigation in information spaces, are also meaningful on the Web space. They are useful navigation cues and powerful ways of conveying complex information via concise representations. Effectively, they play the role of navigational charts, that is, tools that provide users with abstractions of regions of information on the Web. We have implemented the MaGE system to generate maps. During the demo we will show maps in different domains including bibliographic networks.