Abstract
In this research, we aim to automatically investigate local government's involvement in important political tasks by retrieving utterances citing other local government actions from regional assemblies minutes corpus of assembly minutes from all over Japan. To achieve this goal, we should retrieve utterances including local government name first, but there are ambiguous names of local government units.In this paper, we discuss location name disambiguation in the corpus local of assemblies minutes. Our approach to the disambiguation is to collect ambiguous location names from address data and determine correct location name candidate in the utterances including collected ambiguous locations. Determining correct location name candidate utilizes lexical information, local government name of assembly minute, and GIS information. We investigate whether these methods are able to disambiguate location name in high precision.
