医疗器械类拼多多直播带货违法监测预警数据

对采集的拼多多平台带货品类为医疗器械类的达人直播视频内容进行转译分析，对达人口播语言内容进行处理、分析，根据达人在直播过程中对预先设置的违规敏感词（比如：无效退款，保险公司保险，安全，安全无毒副作用，毒副作用小，天然，热销，抢购，试用，家庭必备，免费治疗等）违反的次数和频率，依据触发条件规则提出警告或处理。为诸暨市市场监督局管理区域内规范企业拼多多直播行为，提供数据支持。将采集完成的直播视频进行进行预处理，第一步：基于原始视频文件，以最大10分钟单位对原始视频进行切片。第二步：对于已完成的切片视频，进行视频内容转语音操作。第三步：对于已完成视频转语音操作的切片，进行语音转文本操作。第四步：使用OCR技术对原始视频中抓取的图片进行文字提取操作。第五步：将所得到的文字内容与违法预警关键词库进行匹配。最终运用多标准决策分析模型，对主播在直播过程中出现的违规语句进行分析计算，得出违法预警值和是否预警判断。 违法预警值=（违法预警单关键词命中次数*0.25）+（违法预警组合关键词命中次数* 0.3）+（图片识别命中预警组合关键词个数*0.35）+（直播间近一个月历史违规记录数*0.1） 通过公式计算出最终违法预警值，违法预警值 ≤1 时，不触发预警提示，违法预警值 ＞1 时触发违法预警提示。Conduct transcription and analysis on livestream videos of influencers promoting medical device products on the Pinduoduo platform. Process and analyze the spoken content of these influencers. Issue warnings or take corresponding regulatory actions based on the violation frequency and count against pre-set sensitive prohibited keywords (e.g., "refund without conditions", "insurance underwritten by insurance companies", "safe", "safe with no toxic and side effects", "minor toxic and side effects", "natural", "hot-selling", "flash sale", "trial use", "essential for household use", "free treatment", etc.), in accordance with the preset trigger condition rules. This work provides data support for standardizing the livestream marketing behaviors of enterprises within the jurisdiction of the Zhuji Municipal Market Supervision Administration via the Pinduoduo platform. Preprocess the collected livestream videos in the following steps: Step 1: Segment the original video files into clips with a maximum duration of 10 minutes each. Step 2: Convert the content of the clipped videos into audio. Step 3: Transcribe the converted audio from the clipped videos into text. Step 4: Extract text from images captured in the original videos using OCR technology. Step 5: Match the obtained text content against the prohibited early-warning keyword database. Finally, apply a multi-criteria decision analysis model to analyze and calculate the violation statements made by the influencers during the livestream, and derive the violation early-warning score and the early-warning judgment result. The formula for calculating the violation early-warning score is as follows: Violation Early-warning Score = (Number of hits for single prohibited early-warning keyword * 0.25) + (Number of hits for combined prohibited early-warning keywords * 0.3) + (Number of combined early-warning keywords identified via image recognition * 0.35) + (Number of historical violation records of the livestream room in the past month * 0.1) The final violation early-warning score is calculated using the above formula. No early-warning prompt will be triggered when the score ≤ 1, while an early-warning prompt will be triggered when the score > 1.