Update tagger Makefile with the latest German UD model

Author: AngledLuffa

scripts/pos-tagger/Makefile

@@ -9,7 +9,7 @@ ENGLISH_TEST = /u/nlp/data/pos-tagger/english/test-wsj-19-21
 
 FRENCH_TEST = format=TREES,/u/nlp/data/lexparser/trees/FrenchCC/FTB-Test.utf8.txt
 
-GERMAN_TEST = format=TREES,trf=edu.stanford.nlp.trees.international.negra.NegraPennTreeReaderFactory,/u/nlp/data/GermanACL08/negra/negra_3.mrg
+GERMAN_TEST = format=TSV,wordColumn=1,tagColumn=3,comments=True,/u/nlp/software/CoreNLP-models/de/pos/german-ud/4.3.0/de_gsd-ud-test.conllu
 
 SPANISH_TEST = format=TREES,/u/nlp/data/spanish/ancora/ancora.test,/u/nlp/data/spanish/ldc/ldc-NW.test,/u/nlp/data/spanish/ldc/ldc-DF.test
 
@@ -52,9 +52,10 @@ french.tagger: $$@.props
 	java -mx6g edu.stanford.nlp.tagger.maxent.MaxentTagger -props $@.props > $@.out 2>&1
 	java -mx6g edu.stanford.nlp.tagger.maxent.MaxentTagger -model $@ -testFile $(FRENCH_TEST)  -verboseResults false >> $@.out 2>&1
 
-german: german-dewac.tagger german-fast.tagger german-fast-caseless.tagger german-hgc.tagger
+# TODO: people might still appreciate a fast tagger?
+german: german-ud.tagger
 
-german-dewac.tagger german-fast.tagger german-fast-caseless.tagger german-hgc.tagger: $$@.props
+german-ud.tagger: $$@.props
 	@echo Training $@
 	@echo Will test on $(GERMAN_TEST)
 	java -mx6g edu.stanford.nlp.tagger.maxent.MaxentTagger -props $@.props > $@.out 2>&1

scripts/pos-tagger/german-ud.tagger.props

@@ -0,0 +1,36 @@
+## tagger training invoked at Mon Sep 13 00:20:27 PDT 2021 with arguments:
+                   model = german-ud.tagger
+                    arch = left3words,naacl2003unknowns,unicodeshapes(-2,2),distsim(/u/nlp/data/german/ner/hgc_175m_600,-1,1),distsimconjunction(/u/nlp/data/german/ner/hgc_175m_600,-1,1),unicodeshapeconjunction(-1,1)
+            wordFunction = 
+               trainFile = format=TSV,wordColumn=1,tagColumn=3,comments=True,/u/nlp/software/CoreNLP-models/de/pos/german-ud/4.3.0/de_gsd-ud-train.conllu;format=TSV,wordColumn=1,tagColumn=3,comments=True,/u/nlp/software/CoreNLP-models/de/pos/german-ud/4.3.0/de_gsd-ud-dev.conllu
+         closedClassTags = 
+ closedClassTagThreshold = 40
+ curWordMinFeatureThresh = 2
+                   debug = false
+             debugPrefix = 
+            tagSeparator = _
+                encoding = utf-8
+              iterations = 100
+                    lang = german
+    learnClosedClassTags = false
+        minFeatureThresh = 2
+           openClassTags = 
+rareWordMinFeatureThresh = 10
+          rareWordThresh = 5
+                  search = owlqn2
+                    sgml = false
+            sigmaSquared = 0.0
+                   regL1 = 0.625
+               tagInside = 
+                tokenize = true
+        tokenizerFactory = 
+        tokenizerOptions = asciiQuotes
+                 verbose = false
+          verboseResults = true
+    veryCommonWordThresh = 250
+                xmlInput = null
+              outputFile = 
+            outputFormat = slashTags
+     outputFormatOptions = 
+                nthreads = 1
+        minWordsLockTags = 1