Split recognized text into sentences using punctuation, with timestamps. Args: punc_id_list (Tensor/list): Punctuation IDs from CT-Transformer. Values: 1=none, 2=comma, 3=period, 4=question. timestamp_postprocessed (list): Per-character timestamps [[start_ms, end_ms]
(
punc_id_list, timestamp_postprocessed, text_postprocessed, return_raw_text=False
)
| 123 | |
| 124 | |
| 125 | def timestamp_sentence( |
| 126 | punc_id_list, timestamp_postprocessed, text_postprocessed, return_raw_text=False |
| 127 | ): |
| 128 | |
| 129 | """Split recognized text into sentences using punctuation, with timestamps. |
| 130 | |
| 131 | Args: |
| 132 | punc_id_list (Tensor/list): Punctuation IDs from CT-Transformer. |
| 133 | Values: 1=none, 2=comma, 3=period, 4=question. |
| 134 | timestamp_postprocessed (list): Per-character timestamps [[start_ms, end_ms], ...]. |
| 135 | text_postprocessed (str): Space-separated recognized text. |
| 136 | return_raw_text (bool): Include raw_text in output. |
| 137 | |
| 138 | Returns: |
| 139 | list[dict]: Sentences with keys: text, start, end, timestamp, [raw_text]. |
| 140 | """ |
| 141 | punc_list = ["，", "。", "？", "、"] |
| 142 | res = [] |
| 143 | if text_postprocessed is None: |
| 144 | return res |
| 145 | if timestamp_postprocessed is None: |
| 146 | return res |
| 147 | if len(timestamp_postprocessed) == 0: |
| 148 | return res |
| 149 | if len(text_postprocessed) == 0: |
| 150 | return res |
| 151 | |
| 152 | if punc_id_list is None or len(punc_id_list) == 0: |
| 153 | res.append( |
| 154 | { |
| 155 | "text": text_postprocessed.split(), |
| 156 | "start": timestamp_postprocessed[0][0], |
| 157 | "end": timestamp_postprocessed[-1][1], |
| 158 | "timestamp": timestamp_postprocessed, |
| 159 | } |
| 160 | ) |
| 161 | return res |
| 162 | if len(punc_id_list) != len(timestamp_postprocessed): |
| 163 | logging.warning("length mismatch between punc and timestamp") |
| 164 | sentence_text = "" |
| 165 | sentence_text_seg = "" |
| 166 | ts_list = [] |
| 167 | sentence_start = timestamp_postprocessed[0][0] |
| 168 | sentence_end = timestamp_postprocessed[0][1] |
| 169 | texts = text_postprocessed.split() |
| 170 | punc_stamp_text_list = list( |
| 171 | zip_longest(punc_id_list, timestamp_postprocessed, texts, fillvalue=None) |
| 172 | ) |
| 173 | for punc_stamp_text in punc_stamp_text_list: |
| 174 | punc_id, timestamp, text = punc_stamp_text |
| 175 | if sentence_start is None and timestamp is not None: |
| 176 | sentence_start = timestamp[0] |
| 177 | # sentence_text += text if text is not None else '' |
| 178 | if text is not None: |
| 179 | if "a" <= text[0] <= "z" or "A" <= text[0] <= "Z": |
| 180 | sentence_text += " " + text |
| 181 | elif len(sentence_text) and ( |
| 182 | "a" <= sentence_text[-1] <= "z" or "A" <= sentence_text[-1] <= "Z" |
no outgoing calls
no test coverage detected
searching dependent graphs…