Datasets:
Some of the link are not correct.
Hello, thanks for your work on Cantonese. When I check the data, I have some concerns that will make the cut_point make no sense:
Some of the links direct users to a list of videos. e.g.
{"key": "gd0027722_620270_624350", "rover_result": "但系小母你又睬都唔睬佢入咗船舱添", "confidence": 0.938, "jyutping_confidence": 0.938, "duration": 4.352, "meta_info": {"region": "广东", "program": "封开Boy", "link": "https://www.douyin.com/user/MS4wLjABAAAAHZBNiNY5IdqbOIvfnNgre0HPjf0rrrRO3YdWBf0DpqI"}, "cut_point": [620270, 624350], ...Some of the links are empty. e.g.
{"key": "009608064505_RqohB_8_9080", "rover_result": "趁住李飞分娩嘅时候晕血人事不知将一只狸猫剥咗皮血淋淋咁换走咗啱出世嘅小太子", "confidence": 0.946, "jyutping_confidence": 0.973, "duration": 9.08, "meta_info": {"region": "未知地区", "program": "未知节目", "link": ""}, "cut_point": [73360, 81660], ...
It looks like a common problem since I only check the first 10 objects in the JSONL file. It is hard for others to use the dataset in such situations. Could you fix the error? Or do you plan to release the processed audio file?