[mflare]

Nice work. I also started a very very similar news aggregation service a few months ago, which extracts locations from news articles and shows the headlines on a map. http://mapflare.com

As other comments already mentioned, there are many false positive locations, such as names of persons, organizations, chemical elements or even normal verbs and nouns. There exist for example places named "Robin Hood".

Took me some time to realize that i should not limit the text extraction to locations, but also focus on the recognition of other entities (persons, organizations,..) in order to filter out the ambiguous names.