Merge pull request #2 from Bennycopter/aux-verbs-particles-fix

Use particles instead of aux verbs; ignore some verbs

src/jreadability/jreadability.py

@@ -58,36 +58,37 @@ def split_japanese_sentences(doc: List[UnidicNode]) -> List[List[UnidicNode]]:
 
     mean_length_of_sentence = sum(sentence_lengths) / len(sentences)
 
-    # next, compute proportion of kango, wago, verbs and auxiliary verbs
+    # next, compute proportion of kango, wago, verbs and particles
     num_kango = 0
     num_wago = 0
     num_verbs = 0
-    num_aux_verbs = 0
+    num_particles = 0
     for token in doc:
 
         goshu = token.feature.goshu # goshu (語種) is the word's origin
-        pos = token.feature.pos1
+        pos1 = token.feature.pos1
+        pos2 = token.feature.pos2
 
         if goshu == '漢': # 'kan', meaning chinese
             num_kango += 1
         elif goshu == '和': # 'wa', meaning japanese
             num_wago += 1
 
-        if pos == '動詞': # 'doushi', meaning verb
+        if pos1 == "動詞" and pos2 != "非自立可能": # 'doushi', meaning verb; but not certain verbs like あり in あります
             num_verbs += 1
-        elif pos == '助動詞': # 'jodoushi', meaning auxiliary verb
-            num_aux_verbs += 1
-    
+        elif pos1 == "助詞": # 'joshi', meaning particles
+            num_particles += 1
+
     proportion_of_kango = 100.0 * num_kango / len(doc)
     proportion_of_wago = 100.0 * num_wago / len(doc)
     proportion_of_verbs = 100.0 * num_verbs / len(doc)
-    proportion_of_aux_verbs = 100.0 * num_aux_verbs / len(doc)
+    proportion_of_particles = 100.0 * num_particles / len(doc)
 
     readability_score = mean_length_of_sentence * -0.056 + \
                         proportion_of_kango * -0.126 + \
                         proportion_of_wago * -0.042 + \
                         proportion_of_verbs * -0.145 + \
-                        proportion_of_aux_verbs * -0.044 + \
+                        proportion_of_particles * -0.044 + \
                         11.724
-    
+
     return readability_score