Fix CoreNLP docs: add param help, fold READMEs into tool/DM help, remove stale test-data README, correct language and model-size claims

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: ksuderman

data_managers/data_manager_corenlp_models/README.md

@@ -18,7 +18,8 @@
         #end if
     ]]></command>
     <inputs>
-        <param name="languages" type="select" label="Language Models" multiple="true" display="checkboxes">
+        <param name="languages" type="select" label="Language Models" multiple="true" display="checkboxes"
+               help="Select one or more languages to download. Each model JAR is fetched from Maven Central and registered in the corenlp_models data table. Downloads are large (tens to hundreds of MB each) and may take several minutes.">
             <option value="ar">Arabic</option>
             <option value="zh">Chinese</option>
             <option value="en" selected="true">English</option>
@@ -77,8 +78,8 @@ Usage
 3. Run the data manager
 4. The model JARs will be downloaded from Maven Central and registered in the data table
 
-The language models are large files (typically 100-500 MB each), so downloading
-multiple models may take several minutes depending on your connection speed.
+The language models are large files (from ~80 MB up to several hundred MB each), so
+downloading multiple models may take several minutes depending on your connection speed.
 
 Common Models
 -------------

data_managers/data_manager_corenlp_models/data_manager/data_manager_corenlp_models.xml

@@ -81,8 +81,10 @@ parse.model = edu/stanford/nlp/models/lexparser/italianFactored.ser.gz
 ]]></configfile>
     </configfiles>
     <inputs>
-	    <param name="input" type="data" format="txt" label="Text"/>
-	    <param name="language_model" type="select" label="Language Model">
+	    <param name="input" type="data" format="txt" label="Text"
+	    	help="Plain text file to annotate. Its language must match the selected language model."/>
+	    <param name="language_model" type="select" label="Language Model"
+	    	help="Which installed CoreNLP language model to use. Models are installed by an administrator via the Stanford CoreNLP Language Models data manager (Admin &gt; Local Data). Full annotation is available for English, Chinese, French, German, Italian, and Spanish; Arabic and Hungarian support segmentation only.">
 	    	<options from_data_table="corenlp_models">
 	    		<column name="value" index="0"/>
 	    		<column name="name" index="1"/>
@@ -94,15 +96,17 @@ parse.model = edu/stanford/nlp/models/lexparser/italianFactored.ser.gz
 	    		<filter type="sort_by" column="1"/>
 	    	</options>
 	    </param>
-	    <param name="annotators" type="select" label="Annotation types">
+	    <param name="annotators" type="select" label="Annotation types"
+	    	help="Which CoreNLP annotators to run. Selections are cumulative: each option also runs the annotators it depends on. NER and parsing require an English, Chinese, French, German, Italian, or Spanish model; coreference requires English or Chinese (plus the common models); sentiment requires English.">
 	    	<option value="tokenize" selected="true">Segmentation (sentences and tokens) - All languages</option>
 	    	<option value="pos">Part of speech and lemmas - All languages</option>
 	    	<option value="ner">Named entity recognizer - English, Chinese, French, German, Italian, Spanish</option>
 	    	<option value="parse">Dependency/Constituency parser - English, Chinese, French, German, Italian, Spanish</option>
 	    	<option value="coref">Coreference - English and Chinese only</option>
 	    	<option value="sentiment">Sentiment analysis - English only</option>
 	    </param>
-		<param name="format" type="select" label="Output format">
+		<param name="format" type="select" label="Output format"
+			help="JSON gives the most complete annotation. CoNLL and CoNLL-U are tabular; Text is human-readable; XML is the CoreNLP XML format. Not every annotation is representable in every format (see the help below).">
 			<option value="json" selected="true">JSON</option>
 			<option value="conll">CoNLL</option>
 			<option value="conllu">CoNLL-U</option>
@@ -179,7 +183,7 @@ Part of speech and lemmas
 Named entity recognition (NER)
 	Includes the sentence, token with POS and lemmas, as well as named entities such as PERSON, ORGANIZATION, PLACE, etc.
 
-Dependency parse
+Dependency and constituency parse
 	Includes all of the annotators for NER as well as a dependency parse to
 	find the dependencies between the words and phrases of a sentence in order
 	to determine its grammatical structure.  Some of the tree structures can
@@ -197,6 +201,39 @@ Sentiment analysis
 
     See: https://stanfordnlp.github.io/CoreNLP/sentiment.html
 
+Supported Languages
+===================
+
+Full annotation pipelines (part of speech, NER, parsing, etc.) are available for
+English, Chinese, French, German, Italian, and Spanish. Arabic and Hungarian models
+can be installed but only support segmentation (tokenization and sentence splitting).
+
+Annotator availability by language:
+
+- Segmentation, part of speech, and lemmas: all installed languages
+- Named entity recognition and dependency/constituency parsing: English, Chinese,
+  French, German, Italian, Spanish
+- Coreference: English and Chinese only (requires the common models)
+- Sentiment analysis: English only
+
+Output Formats
+==============
+
+- **JSON** — full annotations with hierarchical structure (recommended)
+- **CoNLL** — tab-separated format for NER and basic annotations
+- **CoNLL-U** — Universal Dependencies format
+- **Text** — human-readable plain text
+- **XML** — CoreNLP structured XML
+
+Installing Language Models
+==========================
+
+Language models are installed by a Galaxy administrator using the **Stanford CoreNLP
+Language Models** data manager (Admin > Local Data): choose the language(s) to install,
+and also install the common models if you need coreference resolution. The tool runs in
+the ``ksuderman/corenlp:@TOOL_VERSION@`` Docker container, which bundles the base CoreNLP
+library; only the language-specific model JARs are installed separately.
+
 **NOTE**
 
 Please note that not all annotation types can be represented in all of the output