Location Extractor

The “Location Extractor” node allows to extract mentionings of geographic locations (aka. toponyms) from unstructured English text.

The location extraction algorithm performs various steps for recognizing potential locations within a given text, followed by a disambiguation. The disambiguation step checks hierarchical/contains relations and identifies the correct locations by their proximity to other given locations in the text (for example, it tries to distinguish between Paris, France and Paris, Texas based on the context in the given text input).

Each identified location in the text is returned, multiple occurrences are returned as often as they occur. Extracted locations are classified into the following categories:

Location Type	Explanation, Example
CONTINENT	e.g. “Asia”
COUNTRY	e.g. “Japan”
CITY	e.g. “Tokyo”
ZIP	Zip code of a city
STREET	Name of a street
STREETNR	Number of a building within a street
UNIT	A political or administrative unit like a federal state, a county, or a city district (e.g. “California”, “Bavaria”, or “Manhattan”)
REGION	An area which is independent from or spanning multiple political or administrative units (e.g. “Midwest”)
POI	A human-made point of interest or a building, like hotels, museums, universities, monuments, etc. (e.g. “Stanford University” or “Tahrir Square”)
LANDMARK	Geographic features like rivers, canyons, lakes, islands, waterfalls, etc. (e.g. “Rocky Mountains”)
UNDETERMINED	An undetermined or unknown type

For each location, geographical coordinates with longitude and latitude values are provided. They are in WGS84 decimal degrees.

In order to use the “Location Extractor”, a “Location Source” (also known as Gazetteer) must be connected to the node’s input port. The Location Source is a database with real world locations and meta information such as alternative names, population counts, coordinates, and hierarchical relations. Currently, there’s the following location sources available:

• The “GeoNames Location Source” connects to the GeoNames REST API.
• The “Local Location Source” is a locally hosted database.

This node uses Palladian’s location extraction mechanism – for more information see: “To Learn or to Rule: Two Approaches for Extracting Geographical Information from Unstructured Text”; Philipp Katz and Alexander Schill; Proc. of the 11th Australasian Data Mining & Analytics Conference (AusDM 2013).