A little over a week ago, I was invited to a classified NGA facility to discuss geospatial technologies and capabilities.
One of the things that came out of the conversation afterward, which I can discuss is AeroText, which is a product developed by Lockheed-Martin. AeroText shares some similarities to MetaCarta, in that it performs natural-language parsing and indexing of unstructured data.
Where AeroText particularly excells is in identifying things like people, things, places, events, temporal references, and co-references, for example if the subject of the discussion is Bill Gates, a subsequent reference to "he" or "his" may then be joined back to Bill Gates, as would other aliases, such as William Gates, Bill, or Mr. Gates.
AeroText appears to be a very powerful tool for intelligent extraction of data, and even supports multilingual applications.
On the geospatial side, the application uses a gazetteer of places of interest for resolving placenames. It appears to be very effective, allowing similar aliases to be utilized, however the content will be limited to locations contained in the gazetteer.
Here MetaCarta has the upper hand, as, in addition to a gazetteer, it also utilizes the Sagent geocoding engine to allow any street address to be resolved.
AeroText nonetheless appears to be a very robust and intriguing tool, with the main driving force being gurus who are linguists, as opposed to geospatial folks. However, I do see potential for tremendous growth for AeroText as it matures.
Subscribe to:
Post Comments
(
Atom
)
0 Response for the " AeroText "
Post a Comment