Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.
Python asraudiochineseemotion-recognitionfunasrmcp-servermultilingual-asropenai-compatible-apiparaformerpunctuationpytorchreal-time-asrspeaker-diarizationspeech-recognitionspeech-to-textstreaming-asrtranscriptionvllmvoice-activity-detectionwhisper-alternative
sensevoice2jsonl.py punctuation matching fix (#2533)
yuGAN63445cd9652c7
diff --git a/funasr/datasets/audio_datasets/sensevoice2jsonl.py b/funasr/datasets/audio_datasets/sensevoice2jsonl.pyindex df6ceaea8..102e82797 100644--- a/funasr/datasets/audio_datasets/sensevoice2jsonl.py+++ b/funasr/datasets/audio_datasets/sensevoice2jsonl.py@@ -4,6 +4,7 @@ import torchimport loggingimport hydraimport re+import stringfrom omegaconf import DictConfig, OmegaConfimport concurrent.futuresimport librosa@@ -119,8 +120,11 @@ def gen_jsonl_from_wav_text_list(dist.barrier()def contains_punctuation(s):- pattern = r'[!"#$%&\'()*+,-./:;<=>?@[\\]^_`{|}~]'- return re.search(pattern, s) is not None+ punctuations = (+ string.punctuation ++ ',。、;:?!""''()【】《》〈〉「」『』〔〕[]{}~·…—–'+ )+ return any(char in punctuations for char in s)def parse_context_length(data_list: list, data_type: str, id=0):pbar = tqdm(total=len(data_list), dynamic_ncols=True)