论人机对抗中“直播+”敏感信息的衍生和把关*

杨 忠

(安徽广播影视职业技术学院 合肥 230011)

On Derivation and Control of "Live +" Sensitive Information in Man-Machine Confrontation

Yang Zhong

(Anhui Broadcasting Movie and Television College,Hefei 230011)

Abstract:[Purpose/Significance]In order to curb the brutal growth of the art and technology of live broadcasting, most platforms currently adopt sensitive information capture mode combining cognitive algorithms and manual auditing, but the impromptu activities and their derived chaotic information still lack in-depth perception. This study initially analyzes the play and mentality behind the communication interface to improve the control strategy. [Method/Process]This paper conducts interdisciplinary exploration according to the methods of functional analysis and behavioral research. The external derivation of live broadcasting sensitive information can be divided into scene dimensions such as launch, fission, fermentation and storage. Experiencing the "net sense" depth during the whole process, having perception of stock width of all information from different means, fully controlling guiding scale and efficiently keeping track of the audio-visual heat, all these multi-dimensions have enriched the visual and even audible perception closed-loop of the three levels from anchor, platform and algorithm. [Result/Conclusion]Live broadcasting external derivative research expands and deepens the observation path for internal control, and coordinates the collaborative blind spot among people, platforms and algorithms. The live broadcast system more clearly standardizes the connotation of the host’s behavior, integrates the platform resources, perfects the algorithm defect, constructs the space-time axis of the live broadcasting and its derivative activities, platforms and intelligent systems that restrict each other and apply scientifically, and promotes the cross-domain perception of the algorithm management and human algorithm from confrontation to unity.

Keywords:sensitive information; man-machine integration; live broadcasting; derivation; intranet

未来网络直播将从纯娱乐转向应用和娱乐并重，行业价值和商业价值凸显。原先的秀场和泛生活直播附加了应用功能，教师、名嘴、网红、村播和官员加盟在线教学、带货和新闻发布会等常态化直播。直播时间长、数量多、范围广、自主性大，稍纵即逝的直播信息还会被品牌生产(BGC)、专业生产(PGC)和用户生产(UGC)叠加再造和重组，使即兴口语表达融入数字域或智能域的时空互动。在人人具有话语权的众创活动中，传受双方的文字、图片和音视频被多语境盘活，用户在“说”和“玩”中参与直播体验，感受价值、渠道和技术的便捷合成。直播活动是网络安全重要关注点，敏感信息过滤应依靠主播自觉、人工审核和智能算法三重维度把关完成。人的感知系统从演播时空拓展到社交时空和抓取时空，从传播艺术转向物质流变。直播发射的敏感信息会形成团体化、组织化和大众化的发酵和裂变，其收集过程源于初始敏感，成于组织监督，归于无缝对接。创意空间内眼、耳、鼻、舌、身都可能是共享吸引子，单介质、网络介质、智能介质共生背景下的直播活动不是孤立存在，任何细微视听都会全程发射、全息裂变、全员发酵和全效收纳。手机变为应用编辑器，算法程序可以根据人声构建面孔，“小度”“小艾”“小雅”“回声”“谷歌之家”“天猫精灵”“AI主播”等虚拟视听具有丰富的敏感性。透过认知和情感的奇异吸引子体验，网络安全系统启动预处理和映射管理收集直播敏感信息，甄别场景形态和分析用户服务要素，并及时发觉、屏蔽、预警、降权乃至封号，管控直播话题的风险导向。……