fix: exclude auxiliary grammar tails from subtitle annotations

src/core/services/tokenizer/annotation-stage.test.ts

@@ -234,6 +234,19 @@ test('shouldExcludeTokenFromSubtitleAnnotations excludes explanatory ending vari
   }
 });
 
+test('shouldExcludeTokenFromSubtitleAnnotations excludes auxiliary-stem そうだ grammar tails', () => {
+  const token = makeToken({
+    surface: 'そうだ',
+    headword: 'そうだ',
+    reading: 'ソウダ',
+    pos1: '名詞|助動詞',
+    pos2: '特殊',
+    pos3: '助動詞語幹',
+  });
+
+  assert.equal(shouldExcludeTokenFromSubtitleAnnotations(token), true);
+});
+
 test('shouldExcludeTokenFromSubtitleAnnotations keeps lexical tokens outside explanatory ending family', () => {
   const token = makeToken({
     surface: '問題',

src/core/services/tokenizer/annotation-stage.ts

@@ -100,6 +100,7 @@ function normalizePos1Tag(pos1: string | undefined): string {
 
 const SUBTITLE_ANNOTATION_EXCLUDED_POS1 = new Set(['感動詞']);
 const SUBTITLE_ANNOTATION_GRAMMAR_ONLY_POS1 = new Set(['助詞', '助動詞', '連体詞']);
+const AUXILIARY_STEM_GRAMMAR_TAIL_POS1 = new Set(['名詞', '助動詞', '助詞']);
 
 function splitNormalizedTagParts(normalizedTag: string): string[] {
   if (!normalizedTag) {
@@ -156,6 +157,16 @@ function isExcludedTrailingParticleMergedToken(token: MergedToken): boolean {
   return trailingPos1.length > 0 && trailingPos1.every((part) => part === '助詞');
 }
 
+function isAuxiliaryStemGrammarTailToken(token: MergedToken): boolean {
+  const pos1Parts = splitNormalizedTagParts(normalizePos1Tag(token.pos1));
+  if (pos1Parts.length === 0 || !pos1Parts.every((part) => AUXILIARY_STEM_GRAMMAR_TAIL_POS1.has(part))) {
+    return false;
+  }
+
+  const pos3Parts = splitNormalizedTagParts(normalizePos2Tag(token.pos3));
+  return pos3Parts.includes('助動詞語幹');
+}
+
 function resolvePos1Exclusions(options: AnnotationStageOptions): ReadonlySet<string> {
   if (options.pos1Exclusions) {
     return options.pos1Exclusions;
@@ -626,6 +637,10 @@ export function shouldExcludeTokenFromSubtitleAnnotations(token: MergedToken): b
     return true;
   }
 
+  if (isAuxiliaryStemGrammarTailToken(token)) {
+    return true;
+  }
+
   if (isExcludedTrailingParticleMergedToken(token)) {
     return true;
   }