feat: Caption enhancements - punctuation and testing (#97)

* Fix caption test - UserWarning for no utterances

* Use punctuated_word for caption if it exists

* WebVTT file starts with first line 'WEBVTT'

Author: jkroll-deepgram

deepgram/extra.py

@@ -50,13 +50,16 @@ def _to_caption(
             utterances = response["results"]["channels"][0]["alternatives"]
         captions = []
         line_counter = 1
+        if format is Caption.WEBVTT:
+            captions.append("WEBVTT")
         for utt_index, utt in enumerate(utterances):
             words = utterances[utt_index]["words"]
+            word_text = "punctuated_word" if "punctuated_word" in words[0] else "word"
             for i in range(0, len(words), line_length):
                 start_time = words[i]["start"]
                 end_index = min(len(words) - 1, i + line_length - 1)
                 end_time = words[end_index]["end"]
-                text = " ".join([w["word"] for w in words[i:end_index + 1]])
+                text = " ".join([w[word_text] for w in words[i:end_index + 1]])
                 separator = "," if format is Caption.SRT else '.'
                 prefix = "" if format is Caption.SRT else "- "
                 caption = (

tests/test_extra.py

@@ -12,7 +12,7 @@
 deepgram = Deepgram(api_key)
 
 MOCK_SRT = "1\n00:00:05,519 --> 00:00:06,019\nYep.\n\n2\n00:00:07,094 --> 00:00:08,615\nI said it before, and I'll say it\n\n3\n00:00:08,615 --> 00:00:09,115\nagain.\n\n4\n00:00:09,974 --> 00:00:11,514\nLife moves pretty fast.\n\n5\n00:00:11,974 --> 00:00:13,654\nYou don't stop and look around once in\n\n6\n00:00:13,654 --> 00:00:15,701\na while. you could miss it."
-MOCK_WEBVTT = "1\n00:00:05.519 --> 00:00:06.019\n- Yep.\n\n2\n00:00:07.094 --> 00:00:08.615\n- I said it before, and I'll say it\n\n3\n00:00:08.615 --> 00:00:09.115\n- again.\n\n4\n00:00:09.974 --> 00:00:11.514\n- Life moves pretty fast.\n\n5\n00:00:11.974 --> 00:00:13.654\n- You don't stop and look around once in\n\n6\n00:00:13.654 --> 00:00:15.701\n- a while. you could miss it."
+MOCK_WEBVTT = "WEBVTT\n\n1\n00:00:05.519 --> 00:00:06.019\n- Yep.\n\n2\n00:00:07.094 --> 00:00:08.615\n- I said it before, and I'll say it\n\n3\n00:00:08.615 --> 00:00:09.115\n- again.\n\n4\n00:00:09.974 --> 00:00:11.514\n- Life moves pretty fast.\n\n5\n00:00:11.974 --> 00:00:13.654\n- You don't stop and look around once in\n\n6\n00:00:13.654 --> 00:00:15.701\n- a while. you could miss it."
 
 """
 Happy case of captions in SRT format.
@@ -32,5 +32,5 @@ def test_get_WebVTT():
 def test_get_SRT_no_utterances():
     response_no_utts = MOCK_RESPONSE
     del response_no_utts["results"]["utterances"]
-    with pytest.raises(AssertionError):
+    with pytest.warns(UserWarning):
         deepgram.extra.to_SRT(response_no_utts)