资源科学 ›› 2019, Vol. 41 ›› Issue (3): 416-429.doi: 10.18402/resci.2019.03.01

• 专栏: 旅游可持续发展 • 上一篇    下一篇

不同来源地旅游者对北京目的地形象感知差异——基于深度学习的Flickr图片分析

邓宁1(), 刘耀芳1, 牛宇1, 计卫星2   

  1. 1. 北京第二外国语学院旅游科学学院,北京 100024
    2. 北京理工大学计算机学院,北京 100081
  • 收稿日期:2018-10-08 修回日期:2019-02-16 出版日期:2019-03-20 发布日期:2019-03-20
  • 作者简介:

    作者简介: 邓宁,男,湖南湘乡人,博士,副教授,硕士生导师,研究方向为目的地数字营销、旅游大数据与人工智能、旅游信息化。E-mail:dengning@bisu.edu.cn

  • 基金资助:
    北京市社会科学基金项目(17JDGLB006); 北京市教委科技项目(SQKM201710031001); 北京第二外国语学院校级科研项目

Different perceptions of Beijing's destination images from tourists: An analysis of Flickr photos based on deep learning method

Ning DENG1(), Yaofang LIU1, Yu NIU1, Weixing JI2   

  1. 1. School of Tourism Science, Beijing International Studies University, Beijing 100024, China
    2. School of Computer Science, Beijing Institute of Technology, Beijing 100081, China
  • Received:2018-10-08 Revised:2019-02-16 Online:2019-03-20 Published:2019-03-20

摘要:

社交网络图片是旅游目的地形象传播的重要载体,基于图片表征内容的营销传播越来越受到旅游目的地营销组织重视,本文选取Flickr上中国港澳台(香港特别行政区、澳门特别行政区、台湾省)、英国和美国旅游者拍摄的北京图片作为研究素材,采用计算机深度学习算法分析图片表征内容,并从认知和情感2个层面分析、比较了不同来源地游客在北京旅游目的地形象感知方面的异同。研究表明,在认知形象方面,入境旅游者均对自然、建筑较为关注,但在文化艺术、人物、食物等具体维度上关注内容不尽相同。在情感形象方面,令人愉快(Pleasant)的、兴奋的(Exciting)是所有入境旅游者表现的主要情感,但中国港澳台和美国旅游者所拍摄图片隐含投射出困倦疲乏的(Sleepy)情感,而英国旅游者拍摄的图片则暗含不安苦恼(Distressing)的情感。本文利用计算机深度学习算法分析海量UGC图片表征内容为目的地形象研究提供了大数据的方法参考。

关键词: 用户生成内容, Flickr图片, 深度学习, 旅游大数据, 目的地形象, 北京

Abstract:

With the rapid development of the Internet, increasingly more tourists use social media to share travel experiences. User generated content (UGC) has become an important source of information for potential tourists, affecting their perception of tourism destination image (TDI) and tourism decision making. Photos, as one of the carrieres of UGC, is an important tool for the spread of TDI across cultures. However, most of the previous photo-based TDI research mainly adopt content or semantic analysis by human, and the number of samples analyzed was limited. When facing large-scale UGC photo collections, more automated analysis methods are needed to improve efficiency. This article introduces the deep learning theory of computer science into TDI for the first time, taking the pictures released by tourists (from Hong Kong, Macao, and Taiwan of China; the United States; the United Kingdom) as the research samples and using machine analysis from millions of destination related photos to represent content and construct the correspondence between destination cognitive and affective images. With regard to the cognitive image, inbound tourists are more concerned about nature and architecture, but they pay different attention to culture, art, people, food and other aspects. Tourists from Hong Kong, Macao, and Taiwan of China are mostly interested in cultural relics, entertainment activities, and food. UK tourists pay attention to facilities and urban environment. US tourists are more likely to photograph people. With regard to the affective image, “exciting” and “pleasant” are the most significant affective elements for inbound tourists. Hong Kong, Macao, Taiwan of China, and US tourists show some sign of sleepiness, and UK tourists show more sign of distress. According to these results, targeted marketing for the main inbound tourist source markets is needed. For tourists from Hong Kong, Macao, and Taiwan of China, it’s essential to increase the cultural content of Beijing, diversify food and Beijing-style entertainment activities. For UK tourists, well-equipped facilities, diverse urban environments, and distinctive buildings are key marketing contents. Natural scenery and life of people are important symbols for attracting US tourists. In summary, this study used computer deep learning methods to analyze large-scale photo data sets, converting pictorial images into text to extract cognitive and affective images. It has both methodological and managerial implications.

Key words: user generated content, Flickr photos, deep learning, tourism big data, destination image, Beijing