Post-Processing with the Named Entity Recognition API
To perform post-processing in an application built on the Named Entity Recognition API, add post-processing tasks to the Named Entity Recognition engine. You can add the tasks:
-
when you create the engine, if you construct the engine by supplying a configuration. When you create a Named Entity Recognition engine from a configuration the configuration automatically adds any post-processing tasks in the configuration to the engine.
See
EdkEngineCreateFromConfigFile
(C API), theEdkEngine
constructor (Java API), orEDKFactory::GetTextExtractionEngine
(.NET API). -
after you create the engine, by calling the appropriate function:
EdkEngineAddPostProcessingTask
in the C API.addPostProcessingTask
in the Java API.
The following functions allow you to specify a minimum score that a match must have for it to return in the results after all post-processing tasks have completed:
EdkEngineSetPostProcessingThreshold
in the C API.setPostProcessingThreshold
in the Java API.
The matches returned by the EdkGetNextMatch
function in the C API, or by iterating over the matches in the Java API, reflect any modifications made by post-processing. If a post-processing task discards a match or its score does not meet the threshold you have specified, it is not returned at all.
If you configure a post-processing task that processes matches en masse, the API does not return matches until all input has been received. This is necessary because an en masse post-processing task requires all of the matches at the same time.
The Named Entity Recognition SDK includes reference documentation for the API. For more information about the SDK, see Named Entity Recognition SDK Package.