ABRIR at NTCIR-9 GeoTime Task Usage of Wikipedia and GeoNames for Handling Named Entity Information

 ABRIR at NTCIR-8 Approach C t ti f N d E tit d t b Construction of Boolean query by using pseudo relevant documents Named entity is more important than others Verb synonym list is used for increasing coverage Combination of Boolean IR model and probabilistic IR model (Okapi) Penalty is applied for documents that don’t satisfy Boolean query Named entity is more important than others  Failure analysis at NTCIR-8 ons ruc on o ame n y a a ase  Usage of Linked Open Data GeoNames : Large-scale database of geographic entities Wikipedia: Information resource for named entity Database construction  Geographic entities from GeoNames and Wikipedia • Add Japanese translation of GeoNames entry by using Wikipedia information [Takenaka, et. al., 2011] Other named entities from Wikipedia • Usage of DBPedia category information to identfy type of named entity (Person, Organization Place and Infrastructure) for making a list Identification of named entities and relationships between these entities and the query is important 「アフリカ(Africa)」⇔「コンゴ民主共和国(Democratic Republic of the Congo)」 Quality of pseudo relevant documents Topic drift Verb synonym expansion Some verbs are not important for Boolean query construction