The need to explicitly georeference (and thus make them inherently geographically searchable) large resource collec- tions such as the Statistical Accounts of Scotland (SAS) which currently only contain implicit georeferences in the form of placennames is becoming more and more urgent. The crucial obvious precondition for successful georeferenc- ing is the recognition of placename occurrences in text [3]. State-of-the-art machine learning systems for the recogni- tion of geographical entities in newswire text achieve an f- score of over 90% [5]. We describe here an experiment with using an o-the-shelf maximum entropy tagger for recognis- ing location names in the SAS, and show that we achieve similar performances. Section 2 brie y describes the data and its annotation, and in Section 3 we present the results of the experiment together with some brief discussion of prob- lems and a comparison with state-of-the-art results. In Sec- tion 4 we discuss some future directions of work.
Search
Reviews
[Write a review of this article]
Find related articles from these CiteULike users
