Legalo

From STLab

Revision as of 21:46, 13 June 2014

Legalo

Path identification algorithm

   algorithm legalo is
       input An RDF graph G from FRED
       output A set S of RDF triples expressing semantic relations between named entities recognized in G
   
       let E the set of named entities in G
       let C a set of named entity pairs
   
       for each named entity e1 in E
           for each named entity e2 in E
               if e1 != e2'
                   add <e1, e2'> to C
   
       let P a set of tuples <e1, e2, sub-graph>
   
       for each pair <e1, e2> in C
           directPaths ← depth first search in G of direct paths between e1 and e2
           if directPaths is not empty
               for each directPath in directPaths
                   add <e1, e2, directPath> to P
           else  
               for each event event in G
                   leftSubGraph ← depth first search in G for finding the shortest direct paths between e1 and event
                   rightSubGraph ← depth first search in G for finding the shortest direct paths between and event and e2
                   if rightSubGraph does not contain agentive roles
                       add <e1, e2, rightSubGraph> to P
       
       let S a set of RDF triples
       
       for each tuple <e1, e2, sub-graph> in P
           property ← generate semantic relation from p
           add RDF triple e1 - property - e2 to S

Generating semantic relations

In order to generate a semantic relation from each sub-graph identified with the legalo algorithm, the following criteria are taken into account:

• a label of a semantic relation is generated by concatenating the properties found in the statements from an event to an object
• some mappings between properties and terms of a semantic relation are performed according to the following criteria
  • vnrole:Beneficiary → for
  • vnrole:Instrument → with
  • vnrole:Destination → to
  • vnrole:Topic → about
• while traversing the statements, also the types of subject nodes are taken into account. Namely, given a subject node its most general FRED type is added into the label of a semantic relation
• repetitions of terms into a label of a semantic relation are avoided, i.e. in case of consecutive properties in a sub-graph such as born born only one is used into a semantic relation