Computer implemented system and method for geographic subject extraction for short text

A computer program product, comprising computer readable instructions stored on a non-transitory computer readable medium for causing a processor to perform operations comprising:

• obtaining one or more documents posted online, and metadata relating to each posted document;
• determining a set of one or more location candidates from parsing text in each document and metadata, wherein each location candidate is a geographic location referred to, or included in the text;
• determining one or more location resolutions for each location candidate, wherein each location resolution comprises geographic location metadata comprising at least a country;
• determining one or more anchor locations from analysing the metadata wherein each anchor location has a single location resolution;
• performing resolution disambiguation on the one or more location resolutions for each location candidate wherein the one or more anchor locations are added to the set of location candidates to be used as additional location candidates, and resolution disambiguation is performed using a distance based scoring function such that a most likely geographic location for a location candidate having two or more location resolutions is the location resolution with a score corresponding to a shortest total distance between the location resolution and each location resolution for each other candidate location including the one or more anchor locations such that anchor resolutions contribute to the scoring for other location resolutions but are not assigned a score themselves;
• and
• reporting the most likely geographic location for each location candidate.

A computing system comprising at least one processor operatively coupled to at least one memory, wherein the processor is configured to perform operations comprising:

• obtaining one or more documents posted online, and metadata relating to each posted document;
• determining a set of one or more location candidates from parsing text in each document and metadata, wherein each location candidate is a geographic location referred to, or included in the text;
• determining one or more location resolutions for each location candidate, wherein each location resolution comprises geographic location metadata comprising at least a country;
• determining one or more anchor locations from analysing the metadata wherein each anchor location has a single location resolution;
• performing resolution disambiguation on the one or more location resolutions for each location candidate wherein the one or more anchor locations are added to the set of location candidates to be used as additional location candidates, and resolution disambiguation is performed using a distance based scoring function such that a most likely geographic location for a location candidate having two or more location resolutions is the location resolution with a score corresponding to a shortest total distance between the location resolution and each location resolution for each other candidate location including the one or more anchor locations such that anchor resolutions contribute to the scoring for other location resolutions but are not assigned a score themselves;
• and
• reporting the most likely geographic location for each location candidate.