Social media data mining has become a prevailing approach to understanding people’s emotional response to, and perception of, public spaces. Drawing on Google Maps Reviews’ text reviews and shared images, this study develops a deep learning framework that combines Transformer BERT and CNN-VGG models to explore public emotions and visual perception of the East Coast Park (ECP) in Singapore. The results show that (1) public emotions of the ECP are predominantly joy, with minor emotional fluctuations in 2020 due to the pandemic lockdown; (2) CNN-VGG image classification is an effective tool to capture people’s visual preferences of public spaces, (3) public emotions across 12 specific public spaces are predominantly joy and neutral, but the results of people’s visual preference exhibited considerable diversity, and (4) image data is mostly associated with joy or neutral emotions, generally indicating users’ satisfaction of ECP public space, and for this reason, there is a lack of sound base to identify environmental factors associated with people’s negative emotions. This study shows that the proposed deep learning framework is effective in using social media data to understand the public's emotional response to and visual preference of urban public spaces.

中文翻译：

---

新加坡东海岸公园的公众情绪和视觉感知：利用社交媒体数据的深度学习方法

社交媒体数据挖掘已成为了解人们对公共空间的情绪反应和感知的流行方法。本研究利用 Google 地图评论的文本评论和共享图像，开发了一个深度学习框架，该框架结合了 Transformer BERT 和 CNN-VGG 模型，以探索新加坡东海岸公园 (ECP) 的公众情绪和视觉感知。结果显示：（1）ECP公众情绪以欢乐为主，2020年受疫情封锁影响情绪波动较小； （2）CNN-VGG图像分类是捕捉人们对公共空间的视觉偏好的有效工具，（3）12个特定公共空间的公众情绪以快乐和中性为主，但人们视觉偏好的结果表现出相当大的多样性，以及（ 4）图像数据多与欢乐或中性情绪相关，一般表明用户对ECP公共空间的满意度，因此缺乏可靠的基础来识别与人们负面情绪相关的环境因素。这项研究表明，所提出的深度学习框架可以有效地利用社交媒体数据来了解公众对城市公共空间的情感反应和视觉偏好。