Merge branch 'eason/main' into SRT_cleanup

Former-commit-id: a9d17c97f798adbdbe29c5cf413d14ff8848236d

SRT.py

@@ -146,7 +146,7 @@ class SRT_script():
         merge_list = []  # a list of indices that should be merged e.g. [[0], [1, 2, 3, 4], [5, 6], [7]]
         sentence = []
         for i, seg in enumerate(self.segments):
-            if seg.source_text[-1] in ['.', '!', '?'] and len(seg.source_text) > 10:
+            if seg.source_text[-1] in ['.', '!', '?'] and len(seg.source_text) > 10 and 'vs.' not in seg.source_text:
                 sentence.append(i)
                 merge_list.append(sentence)
                 sentence = []
@@ -157,7 +157,7 @@ class SRT_script():
         for idx_list in merge_list:
             segments.append(self.merge_segs(idx_list))
 
-        self.segments = segments  # need memory release?
+        self.segments = segments
 
     def remove_trans_punctuation(self):
         """
@@ -417,9 +417,20 @@ class SRT_script():
                 [real_word, pos] = self.get_real_word(word)
                 if not dict.check(word[:pos]):
                     suggest = term_spellDict.suggest(real_word)
-                    if suggest:  # relax spell check
-                        new_word = word.replace(word[:pos], suggest[0])
-                else:
+
+                    if suggest and enchant.utils.levenshtein(word, suggest[0]) < (len(word)+len(suggest[0]))/4:  # relax spell check
+                        
+                        #with open("dislog.log","a") as log:
+                        #    if not os.path.exists("dislog.log"):
+                        #        log.write("word \t suggest \t levenshtein \n")
+                        #    log.write(word + "\t" + suggest[0] + "\t" + str(enchant.utils.levenshtein(word, suggest[0]))+'\n')
+                        print(word + ":" + suggest[0] + ":---:levenshtein:" + str(enchant.utils.levenshtein(word, suggest[0])))
+                        
+                        
+                        new_word = word.replace(word[:pos],suggest[0])
+                    else:
+                        new_word = word
+                else: 
                     new_word = word
                 ready_words[i] = new_word
             seg.source_text = " ".join(ready_words)

pipeline.py

@@ -80,7 +80,7 @@ def get_sources(args, download_path, result_path, video_name):
         audio_path = args.audio_file
         pass
 
-    return audio_path, audio_file, video_path
+    return audio_path, audio_file, video_path, video_name
 
 def get_srt_class(srt_file_en, result_path, video_name, audio_path, audio_file = None, whisper_model = 'base', method = "stable"):
     # Instead of using the script_en variable directly, we'll use script_input
@@ -249,13 +249,13 @@ def main():
     else:
         VIDEO_NAME = args.video_name
 
-    audio_path, audio_file, video_path = get_sources(args, DOWNLOAD_PATH, RESULT_PATH, VIDEO_NAME)
+    audio_path, audio_file, video_path, VIDEO_NAME = get_sources(args, DOWNLOAD_PATH, RESULT_PATH, VIDEO_NAME)
 
     srt_file_en, srt = get_srt_class(args.srt_file, RESULT_PATH, VIDEO_NAME, audio_path, audio_file)
 
     # SRT class preprocess
     srt.form_whole_sentence()
-    # srt.spell_check_term()
+    srt.spell_check_term()
     srt.correct_with_force_term()
     srt.write_srt_file_src(srt_file_en)
     script_input = srt.get_source_only()