珍珠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珍珠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：光滑细腻，天然眩光，孤品，s级，纯天然，925银防过敏，近珠光等）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。"This dataset focuses on the transcription and analysis of live stream video content from Pinduoduo influencers promoting pearl products. The oral speech content of the influencers is processed and analyzed, and warnings or penalties are issued in accordance with trigger condition rules based on the frequency and total count of violations of pre-configured sensitive violation-related keywords (e.g., smooth and delicate, natural glare, exclusive one-of-a-kind item, S-level, 100% natural, 925 silver anti-allergy, near-luster, etc.) during the live streaming process. This work provides data support for the Market Supervision Administration of Zhuji City to standardize the live streaming behaviors of enterprises operating on the Pinduoduo platform within its jurisdiction. First, the collected raw live stream videos are preprocessed through the following steps: 1. Slice the original video files into segments with a maximum duration of 10 minutes each. 2. Convert the sliced video content into audio via video-to-speech conversion. 3. Convert the audio content of the sliced videos into text via speech-to-text conversion. 4. Extract text from images captured in the original videos using OCR technology. 5. Match the obtained text content against the violation warning keyword database. Finally, a multi-criteria decision analysis (MCDA) model is utilized to analyze and calculate the violation-related statements made by the streamer during the live broadcast, so as to obtain the final violation warning score and make a judgment on whether to trigger a warning. The violation warning score is calculated using the following formula: Violation Warning Score = (Number of hits for single violation-related keywords × 0.25) + (Number of hits for combined violation-related keywords × 0.3) + (Number of combined warning keywords detected via image recognition × 0.35) + (Number of historical violation records of the live stream room in the past month × 0.1) Based on the calculated violation warning score, no warning prompt will be triggered when the score ≤ 1, and a violation warning prompt will be triggered when the score > 1.