fix las version of sentence segmenter

src/tokenizer.py

@@ -190,13 +190,14 @@ class SentenceSegmenter:
 
         for article in texts:
             doc = self.nlp(article)
-            for idx, sent in enumerate(doc.sents):
+            last_was_jump = False
 
+            for idx, sent in enumerate(doc.sents):
                 if idx == 0:
                     # Article opener
                     sentence_candidates.append(1)
                     sentence_boundaries.append(1)
-                elif sent.text.endswith("\n"):
+                elif last_was_jump:
                     # Paragraph break candidate
                     sentence_candidates.append(1)
                     sentence_boundaries.append(0)
@@ -204,6 +205,7 @@ class SentenceSegmenter:
                     sentence_candidates.append(0)
                     sentence_boundaries.append(0)
 
+                last_was_jump = sent.text.endswith("\n")
                 sentences.append(sent.text.replace('\n', '').strip())
                 sentence_masking.append(1)