珍珠类快手直播带货违法监测预警数据
收藏资源简介：
对采集的快手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快手直播行为，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 is constructed by collecting and analyzing live stream video content of Kuaishou influencers engaged in pearl-related livestream commerce. We process and analyze the spoken monologue content of these influencers, and issue warnings or take corresponding actions based on trigger rules, according to the frequency and count of violations against pre-configured sensitive violation-related keywords (e.g., "smooth and delicate", "natural glare", "one-of-a-kind item", "S-level", "100% natural", "925 silver allergy-proof", "near pearl luster", etc.) during the live streaming sessions. This work provides data support for the Zhuji Municipal Market Supervision Administration to standardize the livestream commerce behaviors of enterprises within its administrative jurisdiction. The collected live stream videos undergo the following preprocessing steps: 1. Segment the original video files into clips with a maximum duration of 10 minutes; 2. Extract audio content from the segmented video clips; 3. Perform speech-to-text (STT) transcription on the extracted audio to obtain textual transcripts; 4. Extract text from images captured in the original videos using Optical Character Recognition (OCR) technology; 5. Match all obtained textual content against the violation early warning keyword database. Finally, a multi-criteria decision-making (MCDM) model is employed to analyze and calculate the violation statements made by the influencers during the live stream, to derive the final violation early warning score and the early warning judgment result. The violation early warning score is calculated via the following formula: Violation Early Warning Score = (Number of single violation warning keyword hits * 0.25) + (Number of combined violation warning keyword hits * 0.3) + (Number of combined violation warning keywords identified via image recognition * 0.35) + (Number of historical violation records of the live stream room in the past month * 0.1) The early warning judgment rule is as follows: no early warning prompt will be triggered if the calculated violation early warning score ≤ 1, and a violation early warning prompt will be triggered if the score > 1.
