Xu Zhu, Zhang Chuanbin
Department of Humanities, Nanjing Normal University Zhongbei College, Zhenjiang, China.
School of Journalism and Communication, Nanjing Normal University, Nanjing, China.
Front Psychol. 2022 Jul 29;13:959732. doi: 10.3389/fpsyg.2022.959732. eCollection 2022.
The development and application of network media has seriously impacted the social information dissemination environment dominated by traditional media. To break the dissemination barriers encountered by traditional media, this work probes into the dissemination effect of human-computer interactive advertising news. An in-depth analysis of the current dissemination situation of interactive online advertising (IOA) is firstly conducted, and then the methods to effectively guide and manage audience emotions are studied. Finally, an improved LeNet-5 model is established to identify audience emotions. The improvement of the LeNet-5 model in this work is composed of the following four points. (1) The convolution module sets Inception_conv3 and Inception_conv5 are adopted to replace the third convolutional layer Conv3 and the fifth layer Conv5 of the LeNet-5, respectively. (2) The size of the convolution kernel is changed. The original convolution kernel is replaced by two 3 × 3 convolution kernels in the Inception_conv3 and Inception_conv5 module sets. (3) The number of convolution kernels is reasonably changed. (4) The Batch Normalization (BN) layer is used. The experimental results show that interactive advertisements have the better dissemination effects among the audiences with older age, higher education, and in more developed cities. The improved LeNet-5 network can effectively solve the over-fitting and gradient disappearance, with a good robustness. The recognition rate reaches more than 81%, which is higher than the traditional LeNet-5 network by 3%. It can be known that the accuracy of the improved LeNet-5 network image recognition is significantly promoted. This research provides a certain reference for the optimization of news dissemination.
网络媒体的发展与应用严重冲击了以传统媒体为主导的社会信息传播环境。为打破传统媒体所遭遇的传播壁垒，本研究探讨了人机交互广告新闻的传播效果。首先对交互式网络广告（IOA）的当前传播状况进行了深入分析，进而研究了有效引导和管理受众情绪的方法。最后，建立了改进的LeNet-5模型以识别受众情绪。本研究中LeNet-5模型的改进包括以下四点。（1）卷积模块采用Inception_conv3和Inception_conv5集合分别取代LeNet-5的第三卷积层Conv3和第五层Conv5。（2）改变卷积核大小。在Inception_conv3和Inception_conv5模块集合中，将原来的卷积核替换为两个3×3的卷积核。（3）合理改变卷积核数量。（4）使用批量归一化（BN）层。实验结果表明，交互式广告在年龄较大、受教育程度较高以及城市较发达的受众中具有更好的传播效果。改进后的LeNet-5网络能够有效解决过拟合和梯度消失问题，具有良好的鲁棒性。识别率达到81%以上，比传统LeNet-5网络高出3%。可见，改进后的LeNet-5网络图像识别准确率得到显著提升。本研究为新闻传播的优化提供了一定参考。
