misc fixes, achieves 82% accuracy on corpus words

japanese-rules.lex

@@ -1,5 +1,5 @@
 Begin: Root
-Root: DECISION_ROOT PARTICLE_ROOT DEMONSTRATIVE_ROOT CONJUNCTION_ROOT PRENOUNADJECTIVAL_ROOT AUXILIARY_ROOT ADVERB_ROOT YOI_ADJ_ROOT I_ADJ_ROOT VerbRoot INTERJECTION_ROOT NA_ADJ_ROOT TARU_ADJ_ROOT NounRoot PREFIX_ROOT
+Root: DECISION_ROOT PARTICLE_ROOT AUXILIARY_ROOT DEMONSTRATIVE_ROOT CONJUNCTION_ROOT ADVERB_ROOT PRENOUNADJECTIVAL_ROOT YOI_ADJ_ROOT I_ADJ_ROOT VerbRoot INTERJECTION_ROOT NA_ADJ_ROOT TARU_ADJ_ROOT NounRoot PREFIX_ROOT
 
 NounRoot: NUMBER NOUN_ROOT
 VerbRoot: SURU_V_ROOT NonSuruVerbRoot HONORIFIC_REG_VERB
@@ -14,8 +14,8 @@ PREFIX_SUFFIX:
 '' NOUN_ROOT
 
 DECISION_ROOT:
-'' PROBABLY POS:Decision;;BASE:だ
 '' NA_ADJ_NOUN_DESU POS:Decision;;BASE:だ
+な End POS:Decision;;BASE:だ
 
 HONORIFIC_REG_VERB:
 お NonSuruVerbRoot_HONORIFIC

successrate.rb

@@ -25,7 +25,7 @@ def printSuccessRate(inputLines)
 		if total == 0
 			"0 (0%)"
 		else
-			"#{v} (#{(v*100/total)}%)"
+			"#{v} (#{(v*100.0/total).round}%)"
 		end
 	}
 	puts "total w: #{total}"

word-pos-in-corpus.rb

@@ -7,23 +7,27 @@
 word = ARGV[0]
 
 posCounts = {}
-partsOfSpeech.each { |pos| posCounts[pos] = 0 }
+partsOfSpeech.each { |pos| posCounts[pos] = {} }
 File.open("corpus/corpus-allwords-base-pos.txt").each { |line|
 	spl = line.split(" ")
-	baseform = spl[1]
-	if baseform != word
+	conjform = spl[0]
+	if conjform != word
 		next
 	end
+	baseform = spl[1]
 	pos = spl[2]
-	posCounts[pos] += 1
-}
-maxcount = 0
-bestpos = ""
-posCounts.each { |pos,count|
-	if count > maxcount
-		bestpos = pos
-		maxcount = count
+	if !posCounts[pos].include?(baseform)
+		posCounts[pos][baseform] = 0
 	end
+	posCounts[pos][baseform] += 1
+}
+counts = []
+posCounts.each { |pos,bfcount|
+	bfcount.each { |baseform,count|
+		counts.push([count, baseform, pos])
+	}
+}
+counts.sort! {|a,b| b[0] <=> a[0] }
+counts.each { |count, baseform, pos|
+	puts baseform + " " + pos + " " + count.to_s
 }
-puts bestpos
-