基于用户生成数据与自然语言处理的园林植物感知偏好研究——以武汉城市公园为例  

Research on Perceived Preferences for Garden Plants Based on User-Generated Data and Natural Language Processing:A Case Study of Urban Parks in Wuhan

作　　者：苏畅 陈一秀[1] 殷利华 郭诗怡[3,4] SU Chang;CHEN Yixiu;YIN Lihua;GUO Shiyi(College of Architecture and Urban Planning,Huazhong University of Science and Technology,Wuhan 430074;School of Urban Design,Wuhan University,Wuhan 430072)

机构地区：[1]华中科技大学建筑与城市规划学院,武汉430074 [2]华中科技大学湖北省城镇化工程技术研究中心/自然资源部城市仿真重点实验室 [3]武汉大学城市设计学院城市规划系,武汉430072 [4]湖北省人居环境工程技术研究中心

出　　处：《中国园林》2025年第1期125-132,共8页Chinese Landscape Architecture

基　　金：国家自然科学基金项目(52408063,52208083);湖北省自然科学基金项目(2023AFB139);中央高校基本科研业务费专项资金(2042024kf0029)。

摘　　要：城市公园是城市空间的重要部分,在生态、社会和经济等方面至关重要。园林植物作为主要物质空间要素,对公众身心健康有着积极的恢复和促进作用。但基于公众视角与用户生成数据对园林植物感知行为及偏好特征的深入研究较少。以武汉城市公园为对象,按照社交媒体平台优先推荐植物观赏类城市公园的原则筛选,挖掘并解译用户生成数据,借助自然语言处理算法揭示公众对园林植物的感知偏好特征及影响因素。研究表明,公众对植物的感知主要集中在植物种类、季相和植物印象3个方面。其中植物种类感知与季相、感官呈显著正相关,与公园情感得分呈负相关。平台推荐引导的植物感知重点和公众评论中的实际感知基本一致。最后,依据分析结果提出公园季相、观赏错峰、观赏种类创新、观赏植物科教4个方面的植物景观优化建议,为城市公园的园林植物配置优化、公众感知友好型城市公园规划设计等提供研究实证参考。Rapid urbanization has led to mental health problems among residents,and urban green spaces can alleviate these problems,with landscape plants playing a crucial role.Previous studies have mostly been confined to specific plant species or perceptual characteristics and have seldom been carried out from the public's perspective in combination with user-generated data.With the widespread popularization of the Internet,user-generated data and natural language processing methods have provided new avenues for related research.Given the rapid development of urbanization in Wuhan and the diverse types of parks,taking urban parks in Wuhan as the object,this study mined user-generated data and employed natural language processing algorithms to disclose the perceptual preference characteristics and influencing factors of landscape plants,which provides references for the optimization of plant configuration in urban parks.A total of 11 urban parks with high recommendation popularity on Xiaohongshu were selected as research objects,and ornamental recommended plants were extracted according to the recommendation content.The review data about parks were collected from Ctrip and then processed using natural language processing and analysis methods.The jieba library was used to count high-frequency words,and perceptual characteristic categories were established with reference to previous studies.Highfrequency words were expanded using word vectors to construct a perceptual dictionary.Reviews were segmented,and perceptual frequencies and proportions were calculated.The paddle NLP library was used to conduct sentiment analysis on reviews to obtain the comprehensive sentiment score of the parks.The characteristics of plant perception in the reviews are presented as follows:1)Analysis of high-frequency characteristic words related to plants:Among the top 300 high-frequency words,plant-related words accounted for 5.7%.The"plant species"category appeared most frequently,and"cherry blossom"had the highest word frequency.Compared with th

关 键 词：园林植物 风景园林 城市公园 用户生成数据 自然语言处理 景观感知 

分 类 号：S688[农业科学—观赏园艺]