Resolution of Geographical Names in Information Extraction
This study is concerned with using a machine learning method for extracting geographical named entities from the
unstructured text and using a GeoNames database to differentiate between place and non-place references from the
recognized geospatial named entities. It also resolves ambiguous place names (e.g. a place name referring to many places)
into unambiguous place references using contextual clues and other assumptions. The GENIA tagger is used to annotate the
text, and the YamCha is used to train and test the data sets. To evaluate our results, we have also used the CAFETIERE
system on the test data sets. The evaluation process shows that our results have a high recall and has an approximately
equivalent F-measure with the evaluation results confirming the validity of the chosen method.
Keywords: GeoNames, GENIA tagger, YamCha, CAFETIERE system.