OPENNLP-1791: Document the use of ClassPathModelProvider in dev manual (#919)

* OPENNLP-1791: Document the use of ClassPathModelProvider in dev manual
- adds note on JVM argument since J17 and beyond
- improves formatting of all "Note:" text fragments to better highlight helpful context to the reader

(cherry picked from commit 7b362c8)

Author: mawiesne

opennlp-docs/src/docbkx/doccat.xml

@@ -126,7 +126,7 @@ GMDecrease Major acquisitions that have a lower gross margin than the existing n
 GMIncrease The upward movement of gross margin resulted from amounts pursuant to adjustments \
            to obligations towards dealers .]]>
 			</screen>
-			Note: The line breaks marked with a backslash are just inserted for formatting purposes and must not be
+			<emphasis role="strong">Note</emphasis>: The line breaks marked with a backslash are just inserted for formatting purposes and must not be
 			included in the training data.
 		</para>
 		<section id="tools.doccat.training.tool">

opennlp-docs/src/docbkx/langdetect.xml

@@ -141,7 +141,7 @@ lav     Egija Tri-Active procedūru īpaši iesaka izmantot siltākajos gadalaik
 		nedēļu laika Ekonomikas ministrijai, Finanšu ministrijai un Labklājības ministrijai, lai ar vienotu \
 		pozīciju atgrieztos pie jautājuma izskatīšanas.]]>
 			</screen>
-			Note: The line breaks marked with a backslash are just inserted for formatting purposes and must not be
+			<emphasis role="strong">Note</emphasis>: The line breaks marked with a backslash are just inserted for formatting purposes and must not be
 			included in the training data.
 		</para>
 		<section id="tools.langdetect.training.tool">
@@ -153,7 +153,7 @@ lav     Egija Tri-Active procedūru īpaši iesaka izmantot siltākajos gadalaik
 $ bin/opennlp LanguageDetectorTrainer[.leipzig] -model modelFile [-params paramsFile] \
   [-factory factoryName] -data sampleData [-encoding charsetName]]]>
 				</screen>
-				Note: To customize the language detector, extend the class opennlp.tools.langdetect.LanguageDetectorFactory
+				<emphasis role="strong">Note</emphasis>: To customize the language detector, extend the class opennlp.tools.langdetect.LanguageDetectorFactory
 				add it to the classpath and pass it in the -factory argument.
 			</para>
 		</section>

opennlp-docs/src/docbkx/model-loading.xml

@@ -96,6 +96,39 @@ for(ClassPathModelEntry entry : models) {
             </programlisting>
 
         </para>
+        <para>
+          Moreover, certain OpenNLP models can be obtained via a
+          <emphasis>ClassPathModelProvider</emphasis>, such as OpenNLP's
+          built-in <emphasis>DefaultClassPathModelProvider</emphasis> class.
+          It allows direct use of models available under a certain locale, given
+          that those are present in the classpath and can be loaded.
+
+          <programlisting language="java">
+            <![CDATA[
+final ClassPathModelProvider provider = new DefaultClassPathModelProvider(finder, loader);
+// Here: SentenceModel, other model types accordingly
+final SentenceModel sm = provider.load("en", opennlp.tools.models.ModelType.SENTENCE_DETECTOR, SentenceModel.class);
+if(sm != null) {
+  // do something with the (sentence) model
+}]]>
+          </programlisting>
+
+          In the above example, the finder and loader objects can be created or re-used as shown in the
+          previous code example.
+        </para>
+        <para>
+          <emphasis role="strong">Note</emphasis>:
+          When running on Java 17+, the JVM argument
+          
+          <screen>--add-opens java.base/jdk.internal.loader=ALL-UNNAMED</screen>
+
+          may be required. Without this parameter, OpenNLP uses the JVM bootstrap classpath to locate models
+          rather than the UCP class loader.
+
+          For more advanced or non-standard class loading scenarios, using ClassGraph and implementing a
+          custom provider may cover additional cases beyond the default UCP class loader or
+          JVM bootstrap class path behavior.
+        </para>
 	</section>
 
 
@@ -106,7 +139,7 @@ for(ClassPathModelEntry entry : models) {
             we recommend that you have a look at our setup in the <ulink url="https://github.com/apache/opennlp-models">OpenNLP Models
             repository</ulink>. We recommend to bundle one model per JAR file.
 
-            Make sure you add a <emphasis>model.properties</emphasis> file with the following content
+            Make sure you add a <emphasis>model.properties</emphasis> file with the following content:
 
        <programlisting language="java">
                 <![CDATA[

opennlp-docs/src/docbkx/namefinder.xml

@@ -506,7 +506,7 @@ Precision: 0.8005071889818507
 Recall: 0.7450581122145297
 F-Measure: 0.7717879983140168]]>
          </screen>
-			 Note: The command line interface does not support cross evaluation in the current version.
+			 <emphasis role="strong">Note</emphasis>: The command line interface does not support cross evaluation in the current version.
 		</para>
 		</section>
 		<section id="tools.namefind.eval.api">

opennlp-docs/src/docbkx/parser.xml

@@ -200,7 +200,7 @@ $ opennlp ParserTrainer -model en-parser-chunking.bin -parserType CHUNKING \
 		tool replaces the tagger model inside the parser model with a new one. 
 		</para>
 		<para>
-		Note: The original parser model will be overwritten with the new parser model which
+		<emphasis role="strong">Note</emphasis>: The original parser model will be overwritten with the new parser model which
 		contains the replaced tagger model.
         <screen>
 		<![CDATA[