Put some Maui in your UIMA pipelines, if you are crazy enough
Before extracting topics, you have to build a model.
Here the model, aka MauiFilter is a UIMA resource.
Bind it to an MauiTopicExtractorAnnotator analysis engine description and set the parameters for the model.
It is fully parameterizable as expected with Maui.
Then create your Analysis engine from its description and set its parameters.
And of course, extracted topics are added to the CAS.
TL;DR: show me the code
AnalysisEngineDescription aed = createEngineDescription(MauiTopicExtractorAnnotator.class);
createDependencyAndBind(aed, MauiModelResource.PARAM_MODEL(), MauiModelResource.class, "",
// Parameters
MauiModelResource.PARAM_TRAIN_DIR(), "src/main/resources/data/term_assignment/train_fr",
MauiModelResource.PARAM_VOCABULARY(), "src/main/resources/data/vocabularies/agrovoc_fr.rdf.gz",
MauiModelResource.PARAM_FORMAT(), "skos",
MauiModelResource.PARAM_LANGUAGE(), "fr",
MauiModelResource.PARAM_ENCODING(), "UTF-8"
);
AnalysisEngine ae = createEngine(aed,
// Parameters
MauiTopicExtractorAnnotator.PARAM_TEST_DIR(), "src/main/resources/data/term_assignment/test_fr",
MauiTopicExtractorAnnotator.PARAM_FORMAT(), "skos",
MauiTopicExtractorAnnotator.PARAM_LANGUAGE(), "fr",
MauiTopicExtractorAnnotator.PARAM_NBOFTOPICS(), 10,
MauiTopicExtractorAnnotator.PARAM_VOCABULARY(), "src/main/resources/data/vocabularies/agrovoc_fr.rdf.gz"
);
ae.process(ae.newJCas());Despite the fact that the UIMA API is ugly as hell (even with uimaFit), using uimaui is very easy.