大数据环境下的旅游目的地形象研究综述:数据、方法和技术
A Review of Tourist Destination Image in the Big Data Era: Data, Methods, and Technology
收稿日期: 2023-05-12 修回日期: 2023-09-22
基金资助: |
|
Received: 2023-05-12 Revised: 2023-09-22
作者简介 About authors
段锐(1991—),男,山东德州人,北京第二外国语学院旅游科学学院讲师,研究方向:旅游大数据、目的地营销、商务旅游。E-mail:duanrui@bisu.edu.cn;
邹统钎(1964—),男,江西吉安人,博士,北京第二外国语学院中国文化和旅游产业研究院教授,研究方向:旅游目的地管理与旅游规划、文化遗产管理、文化旅游发展政策。
梁未哲(2000—),女,山西忻州人,北京第二外国语学院旅游科学学院硕士研究生,研究方向:目的地营销。
本文通过收集和筛选发表在科学引文索引(SCI)、社会科学引文索引(SSCI)和北大中文核心期刊目录(2020版)中与大数据环境下旅游目的地形象研究相关的文献,从数据(来源和模态)、方法和技术(工具)3个视角分析总结现有研究,展望未来研究趋势。研究显示:用户生成内容是主要的数据来源,文本模态数据占据主导地位;内容分析法是大数据环境下旅游目的地形象研究的基本方法,情感分析、主题建模、图像标注、图像识别等其他常用的计算机科学领域方法亦可归入此类;技术上,主要通过自然语言处理技术对文本进行分析,深度学习技术逐渐应用于图片和视频数据分析领域;工具上,主要使用带有图形用户界面的数据分析软件进行分析以及通过编程调用平台或模块开展研究。
关键词:
By collecting and screening 74 papers related to tourist destination image research in the big data environment — and published in the Science Citation Index (SCI), Social Science Citation Index (SSCI), and Peking University Chinese Core Journal Catalog (2020 Edition), as of April 2023 — this study combs through the current status of the field, and analyzes and summarizes existing research from three perspectives — data (sources and modalities), methodology, and technology (tools) — and looks forward to future research trends. The research shows that user-generated content is the main data source, textual data is dominant in terms of modality, and content analysis is the basic method of tourist destination image research in the big data environment. Other commonly used methods in the field of computer science — such as sentiment analysis, topic modeling, image annotation, and image recognition — can also be classified into this category. Technically, application programming interfaces (APIs) and software tools are mostly used, and the application of deep learning mainly focuses on image data, which has not yet been fully investigated in research on the image of tourist destinations.
Keywords:
本文引用格式
段锐, 邹统钎, 梁未哲.
DUAN Rui, ZOU Tongqian, LIANG Weizhe.
观点与创新:
● 从数据、方法和技术3个视角综述大数据环境下的目的地形象研究
● “用户生成内容”是大数据环境下目的地形象研究主要的数据来源
● 内容分析法是大数据环境下目的地形象研究的基本方法
● 未来研究应关注新数据来源(人工智能生成内容)、新数据模态(视觉模态)、新技术方法(多模态学习)的应用
Highlights:
● Overview of destination image research in the big data environment from three perspectives: data, methods, and technology.
● User-generated content (UGC) is the main data source for destination image research in big data environments.
● Content analysis is the fundamental method for destination image research in the big data environment.
● Future research should focus on the application of new data sources (AI-generated content), data modalities (visual modalities), and technological approaches (multimodal learning).
引言
旅游目的地形象(tourism destination image,简称TDI)在目的地营销的研究和实践中均发挥重要作用,是品牌管理理论和方法在目的地营销领域的具体应用(邹统钎,2019)。20世纪70年代,Gunn(1972)、Mayo(1973)和Hunt(1975)3位学者最早开展与目的地形象相关的研究(Pike,2002)。由于目的地形象与旅游消费者行为、目的地营销等研究关系密切,经过半个世纪的发展,旅游目的地形象已经成为旅游研究中的热门领域。通过对不同时期目的地形象研究的梳理,可以发现相关研究主要聚焦在4个方向:(1)目的地形象的概念内涵。明确定义目的地形象比较困难,这与目的地形象复杂的心理学和哲学渊源有关,Lai和Li(2016)认为,由于心理学和哲学研究存在不同流派,因此以它们为理论基础的目的地形象概念内涵具有一定的模糊性。(2)目的地形象的内部结构。考虑到直接定义目的地形象的难度,许多研究尝试分析目的地形象的内部结构,并提出了几类模型,目前已在目的地形象研究领域普遍使用,如三连续体模型(Echtner & Ritchie,1991)、认知 — 情感 — 意动维度模型(Gartner,1994)、认知 — 情感 — 整体维度模型(Baloglu & McCleary,1999)、特征模型(Gallarza,Saura & García,2002)、长尾模型(Pan & Li,2011;Stepchenkova & Li,2012)、核心—边缘模型(Lai & Li,2012)等。这些模型从不同视角分析了目的地形象的构成元素、特征和结构,对于理解目的地形象的概念内涵和形成过程,以及度量目的地形象具有指导作用。(3)目的地形象的形成过程。目的地形象的形成经历原生形象(organic image)、引致形象(induced image)和游后形象(post travel image)3个阶段(Gunn,1972),Gartner(1994)进一步细化并汇总了目的地形象形成过程中的8种动因。另外,从外部看,个人因素(包含心理动机和社会人口学因素)和刺激因素(包含各种信息源)影响游客的认知形象和情感形象的形成;从内部看,游客的认知形象影响情感形象,两者共同决定了游客对目的地整体形象的感知(Baloglu & McCleary,1999)。(4)目的地形象的度量。Echtner和Ritchie(1991)认为目的地形象的度量方法有结构化和非结构化两类:从研究方法的视角,前者属于定量研究,后者属于定性研究;从哲学方法论的视角,前者是还原论(Reductionism),后者是整体论(Holism)。
在旅游营销实践中,目的地形象发挥着重要的作用。目的地营销组织(Destination Marketing Organization,简称DMO)通常会精心设计和维护目的地形象,将其作为一种营销和品牌管理策略,并采取有效方式进行传播。在传统媒体时代,目的地形象的传播是单向的,DMO通过电视、报纸、杂志、宣传册等方式宣传目的地形象,影响游客对目的地形象的感知,引致形象产生;在旅游过程中,游客的感知形象可以进一步通过“凝视”得到巩固或重塑,完成一个“诠释循环”的过程(hermeneutic circle)(Urry & Larsen,2011)。随着互联网的发展,尤其是社交媒体的出现,目的地形象的传播过程发生了根本性的变化——从单向传播变为双向互动。在Web 2.0时代,每个用户作为“自媒体”将海量的文本、图片、视频等数据上传到互联网和社交媒体,从信息的被动接收者变为主动创造者。海量数据形成的大数据环境从根本上改变了目的地形象的传播方式,为目的地营销的研究者和从业者带来了新的机遇和挑战。
与产生于科学研究中的学术概念不同,“大数据”(big data)这一概念来自工业界,最早由思科(Cisco)、IBM、甲骨文(Oracle)等信息科技公司提出(Logica & Magdalena,2015),并被高德纳和麦肯锡等咨询公司用于其报告之中(Sagiroglu & Sinanc,2013),从而被人们所熟知。大数据中的“大”字能够给人们带来一种关于数据体量增长的直观感觉并提升其吸引力,但“大”作为一个表示相对意义的形容词,难以给出一个准确的定义,因此大数据并不是一个规范的学术概念。目前普遍被人们所接受的观点是从数据特征上定义“大”,即大数据的“4V”特征:海量性(volume)、多样性(variety)、快速的数据处理(velocity)、价值密度低(value)。另外,目前在学术界提及大数据,并不仅指数据方面的特征,同时还包含了另外两个方面的含义:一是指应用于数据获取、储存、管理、分析过程中的,以机器学习(包含深度学习)、数据挖掘、人工智能等技术为代表的计算机和信息科学技术,统称为“大数据技术”;二是指相对于基于小样本数据,更加接近于基于全样本数据的研究方法,被称为“大数据研究方法”。因此,大数据的概念同时具有数据、技术和方法3个层面的含义。
近十年来,许多学者在大数据背景下开展目的地形象研究。根据前面对大数据概念的分析,大数据给目的地形象研究带来的变革也应该相应地从3个层面来看待:数据层面,存在于互联网、社交媒体的多源异构数据,包含用户生成内容(user generated content,简称UGC)、职业生成内容(occupationally generated content,简称OGC)①(①部分文献中也称“专业生成内容”(professionally generated content,简称PGC)。)以及即将到来的人工智能生成内容(artificial intelligence generated content,简称AIGC),为目的地形象研究提供了更加丰富的数据基础;方法层面,大数据对传统的目的地形象研究中基于定性和定量分析划分的研究范式提出新的挑战,赋予了内容分析法(content analysis)新的含义,引入了诸如主题建模(topic modelling)、情感分析(sentiment analysis)、图像识别(image recognition)、图像描述(image caption)等计算机和信息科学领域的研究方法,为目的地形象研究开辟了新思路;技术层面,以机器学习(尤其是深度学习)为代表的大数据技术,为多模态数据(文本、图片、视频等)自动化处理、分析提供了解决方案。
本文通过收集整理在大数据背景下开展目的地形象研究的国内外相关文献,对数据、技术和方法3种视角的研究现状进行分析,并在此基础上进行总结,指出未来研究的发展趋势。本研究具有一定的创新性,对理论研究和实践应用有指导意义,具体表现为:(1)本文首次尝试对大数据与目的地形象的交叉研究做体系化的梳理分析和总结展望;(2)本文在对当前研究总结和未来展望的基础上,建立了大数据环境下目的地形象研究的基本框架,为相关研究者提供了清晰的研究路径;(3)本文从数据、方法和技术(工具)3个角度对大数据环境下的目的地形象研究进行分析总结,紧密贴合目的地营销实践,具有较强的可操作性,可为实践应用提供方法论指导。
一、文献梳理
为更加全面地收集到与“大数据背景下旅游目的地形象研究”主题相关的论文,本文首先以“旅游目的地形象”“tourism destination image”“大数据”“big data”“用户生成内容”“user generated content”“在线评论”“online review”“博客”“blog”“旅游图片”“tourism photo/picture”“视频”“video”等为中英文关键词,在中国知网、Web of Science、ScienceDirect、Taylor & Francis、Springer、Google Scholar等国内外大型数据库中进行主题检索,论文检索的时间范围选取2023年4月30日前;其次,通过人工筛选保留了收录于SCI、SSCI和北大中文核心期刊目录(2020版)中的文献;最后,通过人工逐一检查,确保每篇论文都与本文所关注的研究主题相关。尤其需要指出的是,按照上文对大数据概念的分析,我们在筛选大数据环境下的目的地形象研究文献时,数据量并不是筛选的唯一标准,因为单独从数据量标准来判断,几乎所有的研究都不能够称为“大”数据研究——数据量到PB级(petabyte)以上(涂新莉、刘波、林伟伟,2014)。为此,我们通过对收集到的128篇文献进行逐一考察,综合考虑数据、方法和技术上的特点,将其中的74篇文献纳入本文的研究范围,如表1所示。
表1 纳入本文研究范围的相关论文
Tab. 1
大数据层面 Level of big data | 细分要素 Subdivision element | 相关论文编号 Related literature number | |
---|---|---|---|
数据 | 来源 | UGC | 1-74 |
OGC | 6,12,13,15,17,19,25,30,49,72 | ||
模态 | 文本 | 1,2,5-7,10,11,13-39,41-47,51-55,57-69,71-74 | |
图片(元数据) | 7-9,50,51 | ||
图片 | 3,4,12,13,40,41,48,56,70 | ||
视频 | 49 | ||
方法 | 内容分析 | 文本挖掘 | 1,6,8,9,10,11,14-22,25-32,39,43-48,50-55,57,58,61-67,69,71-74 |
主题建模 | 2,4,5,9,14,23,24,33-38,41,42,44,59,68 | ||
情感分析 | 2,5,7,8,11-13,17,18,21,23,26,27,30,31,33,34,41,42,45-48,50,51,52,53-55,57,58,60-66,69,71-74 | ||
图片描述 | 3,4,12,13,41 | ||
图像识别 | 40,70 | ||
视频描述 | 49 | ||
技术/工具 | 技术 | 自然语言处理技术 | 1,6,8,9,11,14,17-20,25-27,29,31,32,39,43,50,61,64,67 |
LDA主题建模技术 | 2,5,14,23,24,33-35,37,38,41,42,59,68 | ||
基于情感词典的情感分析技术 | 5,7,8,9,11,17,18,33,34,50 | ||
深度学习技术 | 40,41,56,60,70 | ||
工具 | 带有图形用户界面的 数据分析软件 | 1,6,11-13,15,16,21-23,28,36,41,44-47,51-55,57,58,62,63,65,66,69,71-74 | |
通过Python等编程语言 调用平台或分析模块 | 2,3,4,8,9,10,12,13,41,48,49,61 |
注:
1.由于篇幅所限,
2.“图片(元数据)”指基于图片的元数据(如标题、标签等)对图片进行间接分析的方法,本质上是一种文本分析方法,在本文第二部分有详细介绍。此处单独列出,主要和对图片内容进行直接分析的方法进行区分。
3.“文本挖掘”主要指对文本进行基础的自然语言处理(如去停用词、分词、词干提取、词性标注等)的方法,与后面的“自然语言处理技术”相对应。
4.“深度学习技术”主要包含残差神经网络、卷积神经网络、循环神经网络、长短期记忆网络等技术
经梳理,74篇文献在数据、方法和技术上具有以下特点:
(1)数据方面。全部以UGC为数据来源,少数包含OGC(10篇),主要用于和UGC数据进行对比分析;在数据模态上,以文本数据为主(63篇),图片数据次之(14篇),视频数据目前还极少(仅有1篇)。
(2)方法方面。以内容分析法为主,其中包含文本挖掘、情感分析、主题建模、图像(视频)描述、图像识别(包含人物识别、物体识别、场景识别等)等方法。我们认为目前的大数据研究方法其本质就是一种针对二手数据的内容分析法。
(3)技术方面。以基本的自然语言处理(包含分词、词性标注、词频统计、共现分析等)、基于隐含狄利克雷分布(Latent Dirichlet Allocation,简称LDA)的主题建模、基于情感词典的情感分析(如SentiWordNet、HowNet等)、深度学习(卷积神经网络、残差神经网络、长短期记忆网络)等技术为主。在技术的实现上,大多依赖带有图形用户界面的数据分析软件(如ROST CM、CATPAC II、NVivo、AntConc、ATLAS. ti等),或通过Python等编程语言链接平台接口(application programming interface,简称API)调用相关的开源工具(如Google Cloud Vision、百度开源图像识别)、数据分析模块(如TextBlob、VADER、DeepSentiBank、SnowNLP等)开展研究。
另外,我们对所收集文献的年份(图1)和期刊分布情况进行了统计(表2)。从年份分布看,除了最早的2篇分别发表于2007年和2011年之外,其余论文主要集中在近十年,尤其是在2018年以后,大数据环境下的目的地形象研究在数量上有了快速的增长,这与大数据研究方法和技术的逐渐成熟密切相关。从期刊分布看,SCI和SSCI期刊方面,旅游领域顶级期刊Tourism Management刊发该领域研究论文数量最多(9篇),而Journal of Travel Research仅刊发1篇,Annals of Tourism Research未刊发相关论文。其他刊发相关论文较多的英文期刊还有Sustainability(6篇)和Journal of Destination Marketing & Management(4篇)。中文核心期刊方面,《旅游学刊》刊发相关论文3篇,在数量上与《资源开发与市场》并列第一。
图 1
图 1
大数据环境下旅游目的地形象研究相关论文发表年份分布
Fig. 1
Year distribution of literature related to tourism destination image research under the big data environment
表 2 大数据环境下旅游目的地形象研究相关论文发表期刊分布
Tab. 2
排名 Rank | 期刊 Journal | 论文篇数 Number of literature | 占比(%) Percentage |
---|---|---|---|
1 | Tourism Management | 9 | 12.16 |
2 | Sustainability | 6 | 8.11 |
3 | Journal of Destination Marketing & Management | 4 | 5.41 |
4 | 《旅游学刊》 | 3 | 4.05 |
4 | 《资源开发与市场》 | 3 | 4.05 |
6 | Asia Pacific Journal of Tourism Research | 2 | 2.70 |
6 | Information Technology & Tourism | 2 | 2.70 |
6 | Tourism Management Perspectives | 2 | 2.70 |
6 | 《地域研究与开发》 | 2 | 2.70 |
6 | 《华中师范大学学报》 | 2 | 2.70 |
6 | 《经济地理》 | 2 | 2.70 |
6 | 《旅游科学》 | 2 | 2.70 |
6 | 《资源科学》 | 2 | 2.70 |
14 | African and Asian Studies | 1 | 1.35 |
14 | Asia Pacific Journal of Marketing and Logistics | 1 | 1.35 |
14 | Computer Journal | 1 | 1.35 |
14 | Frontiers of Information Technology & Electronic Engineering | 1 | 1.35 |
14 | Industrial Management & Data Systems | 1 | 1.35 |
14 | International Journal of Environmental Research and Public Health | 1 | 1.35 |
14 | Island Studies Journal | 1 | 1.35 |
14 | Isprs International Journal of Geo-Information | 1 | 1.35 |
14 | Journal of Business Research | 1 | 1.35 |
14 | Journal of Electrical and Computer Engineering | 1 | 1.35 |
14 | Journal of Hospitality & Tourism Research | 1 | 1.35 |
14 | Journal of Travel & Tourism Marketing | 1 | 1.35 |
14 | Journal of Travel Research | 1 | 1.35 |
14 | Kybernetes | 1 | 1.35 |
14 | Soft Computing | 1 | 1.35 |
14 | Tourism Review | 1 | 1.35 |
14 | Tourist Studies | 1 | 1.35 |
14 | 《北京大学学报》 | 1 | 1.35 |
14 | 《出版科学》 | 1 | 1.35 |
14 | 《地理科学进展》 | 1 | 1.35 |
14 | 《地理研究》 | 1 | 1.35 |
14 | 《福建师范大学学报》 | 1 | 1.35 |
14 | 《干旱区资源与环境》 | 1 | 1.35 |
14 | 《桂林理工大学学报》 | 1 | 1.35 |
14 | 《国土资源科技管理》 | 1 | 1.35 |
14 | 《海南大学学报》 | 1 | 1.35 |
14 | 《林业经济问题》 | 1 | 1.35 |
14 | 《农业经济》 | 1 | 1.35 |
14 | 《西北民族大学学报》 | 1 | 1.35 |
14 | 《浙江农林大学学报》 | 1 | 1.35 |
14 | 《中国博物馆》 | 1 | 1.35 |
14 | 《中国岩溶》 | 1 | 1.35 |
14 | 《中南林业科技大学学报》 | 1 | 1.35 |
二、大数据环境下的旅游目的地形象研究:数据视角
从数据视角来看,用于目的地形象研究的大数据呈现出多源异构的特点,“多源”指数据来源丰富,“异构”指数据模态结构各异,这两点是大数据的基本特征(涂新莉、刘波、林伟伟,2014)。
1.数据来源
目前用于目的地形象研究的大数据主要来源于互联网。相较于Web端数据,由于移动端数据便于获取地理位置等信息,目前移动端大数据更加受到研究人员的重视(Lu & Stepchenkova,2015)。具体来说,目前存在于互联网的可用于目的地形象研究的数据类型主要包括UGC和OGC两种。UGC是Web 2.0时代的产物,来源十分广泛,存在于社交媒体、OTA、口碑网站、百科等多种平台。相对于UGC,OGC是一种“专业”内容,如职业摄影师拍摄的目的地宣传照片或视频,专业撰稿人撰写的旅游报道、编写的旅游故事等。UGC和OGC之间的差异主要体现在两个方面:一是立场不同,UGC作为一种来自游客的内容,体现的是需方立场,而OGC作为一种来自目的地营销组织(Destination Marketing Organization,简称DMO)的内容,体现的是供方立场,这也就决定了UGC可用于分析游客对目的地的接收形象(Received Image),而OGC则只能用于研究目的地投射形象(Projected Image);二是作用不同,UGC作为自媒体的产物,也可以发挥投射形象的作用,但其在目的地形象的形成过程中发挥的是原生作用,而OGC作为大众媒体的产物,创作者是既得利益者,在目的地形象的形成过程中发挥的是引致作用。在大数据环境下的目的地形象研究中,UGC是目前应用最普遍的一种大数据。从数据体量上看,OGC很难被称为大数据,所以在目前的研究中,OGC主要用于和UGC进行对比分析。
表3 大数据环境下UGC在目的地形象研究中的应用分析
Tab. 3
UGC作用 Role of UGC | 研究主题 Research topic | 相关论文编号 Related literature number |
---|---|---|
接收形象 Received image | 游客感知形象度量方法设计 | 1-5,14,23,24,26-28,34,37,38,40,41,73 |
分析特定细分市场中游客的感知形象 | 9,31,32,35,39,42-45,50-52,54-67,69,71,74 | |
对比不同游客感知形象的差异 | 8,48,18,29,46,53 | |
分析投射形象和接收形象之间的一致性及其对用户参与的影响 | 7,12,13,17,25,30,72 | |
分析不同阶段游客感知形象的演变 | 16,22,47,68,70 | |
分析外生因素对游客感知形象的影响 | 21,47,68 | |
投射形象 Projected image | 将UGC和OGC的投射形象进行对比分析 | 6,15,49 |
分析UGC在目的地形象形成过程中所发挥的作用 | 10,11,20,32,36 |
(1)游客感知形象度量方法设计。基于UGC的“大数据”特点,利用机器学习领域的模型和技术,设计度量游客感知形象的方法,包含针对文本的自然语言处理(Marine-Roig,2017;Alarcón-Urbistondo,Rojas-de-Gracia & Casado-Molina,2023)、主题建模(Wang,Li & Wu,et al.,2021;Luo,He & Yang,et al.,2023)、情感分析(Cao,Liang & Li,et al.,2020;Ali,Marc & Omar,et al.,2021)方法,以及针对图片的视觉分析方法(Zhang,Chen & Li,2019;Arabadzhyan,Figini & Vici,2021;Arefieva,Egger & Yu,2021;Zhang,Lu & Zhou,et al.,2022)。另外,张珍珍和李君轶(2014)还对比了“小”数据(问卷调查)和“大”数据(网络文本)在目的地形象研究中的异同。
(2)分析特定细分市场中游客的感知形象。这方面的研究以发表在中文期刊的居多,针对某一特定细分市场,分析研究该细分市场内的游客感知形象。在不同研究中,旅游市场的细分方式有所不同,主要有以下几种方式:
③以旅游业态细分,共有8篇文献,研究范围涉及黑色旅游、红色旅游、乡村旅游、冰雪旅游、美食旅游、酒店旅游等诸多业态。如柴海燕、王璐和王思远(2020)针对黑色旅游这一特殊旅游形式,以侵华日军第七三一部队罪证陈列馆为研究对象,通过分析相关旅游网站的评论数据,构建了游客对黑色旅游目的地的形象感知模型。
(4)分析投射形象和接收形象之间的一致性及其对用户参与的影响。将OGC作为目的地投射形象的数据来源,UGC作为游客接收形象的来源,分析两者之间的一致性程度。Költringer和Dickinger(2015)通过收集对比不同来源的在线信息,其中既包含UGC,也包含新闻报道、官方网站等OGC,借此分析了目的地品牌识别的传播;Marine-Roig和Ferrer-Rosell(2018)使用成分分析法,测量了加泰罗尼亚(Catalonia)投射形象和游客感知形象之间的差异;Marchi 和Raschi(2022)以意大利的一个小岛(Capraia)为研究对象,发现了投射形象和感知形象之间的差异——DMO更加注重传播常规的信息,而游客更加重视情感体验;Huang、Han和Meng等(2022)以广西巴马为研究对象,使用重要性 — 表现分析(Importance Performance Analysis)矩阵从视觉上揭示了DMO投射形象和游客感知形象的差异;张文亭和骆培聪(2017)则以福建永定土楼为研究对象,分析了官方网站传播的形象和游客感知形象之间的差异。还有研究进一步分析了投射形象和感知形象的差异对游客行为的影响,He、Deng和Li等(2022)通过深度学习和网络分析方法识别了目的地形象的“认知 — 情感”元素并构建了目的地形象的“核心 — 边缘”结构,分析投射形象和感知形象的差异,在此基础上分析了其对用户评论和点赞的影响,给出了指导DMO选择投射内容的相关策略;Huang、Han和Meng等(2022)分析了投射形象和接收形象之间的差异性对用户体验评价的影响,以及其他3个因素(外部危机、促销内容的可读性、游客的专业性)在其中所发挥的中介作用,同样为DMO的营销策略选择提供了指导;Deng 和Li(2018)则给出一种提高投射形象和接收形象一致性的方法,即通过对UGC图片中认知和情感要素的分析,设计了一种基于朴素贝叶斯模型的机器学习算法来指导DMO图片的选择。
(5)分析不同阶段游客感知形象的演变。Jani和Hwang(2011)以坦桑尼亚的桑给巴尔岛(Zanzibar)为研究对象,比较了游客游前和游后感知形象的不同,发现游前游客更加关注基于功能属性的形象,而游后游客更加关注基于心理属性的形象,且游后游客相比于游前游客有更加正面积极的形象感知;Liu、Liu和Mo等(2020)分析了2014年至2018年游客对澳门感知形象的演变,发现与赌场相关的形象感知在下降,“文化之都”和“世界旅游休闲中心”的形象正在获得游客的认可;张坤、李春林和张津沂(2020)对2004年至2013年十年间入境游客对北京的整体形象感知和行为足迹的时空演变进行了分析;丛丽、徐琳琳和方小雨(2021)分析了北京冬奥会申办前后游客感知形象的变化;徐琳琳、周彬和虞虎等(2023)则分析了在北京冬奥会申办期、筹备期和预热举办期3个阶段游客对张家口感知形象的时序变化。
(6)分析外生因素对游客感知形象的影响。Li、Xu和Song等(2020)以成都为研究对象,分析了抖音上的美食短视频对游客认知、情感、意动形象的影响,为DMO进行目的地形象构建和城市品牌传播提供了借鉴。丛丽、徐琳琳和方小雨(2021)以及徐琳琳、周彬和虞虎等(2023)从“认知—情感—整体”形象的维度出发,分别分析了北京冬奥会的成功申办与举办对北京和张家口城市形象的影响;冬奥会的成功申办使得“冬奥”元素在北京城市形象中得到凸显,且申办前后游客的认知形象维度有所变化,情感形象以积极情感为主;同时,冬奥会的申办和举办使得张家口的城市整体形象日益多元化,冰雪旅游形象相关的认知要素得到加强,情感形象变得更加积极。
UGC作为投射形象的数据来源,在目的地形象的形成过程中发挥着原生动机(organic agent)的作用(Gartner,1994),相关研究较少(9篇),主要聚焦于如下两点。
(1)将UGC和OGC的投射形象进行对比分析。Choi、Lehto和Morrison(2007)较早开展了目的地形象大数据研究,收集了不同来源的OGC和UGC数据,并采用文本挖掘、专家判断和贡献分析等定性和定量结合的方法,对比分析了OGC和UGC投射形象的差异,以及导致这种差异的原因。Iordanova和 Stainton(2019)以柬埔寨为研究对象,对比分析了官方旅游网站和业余旅游博客所发布的关于柬埔寨旅游信息及投射形象的差异,发现官方形象以认知元素为主,而旅游博客则兼具认知和情感元素,更为重要的是在游客感知信任方面,官方内容与旅游博客发挥的作用有明显差距。邓宁和蘧浪浪(2022)通过对YouTube视频的分析,比较了DMO和国外游客在北京目的地形象建构上的异同,并借鉴UGC视频在形象塑造上的成功经验指导DMO开展目的地营销。
(2)分析UGC在目的地形象形成过程中发挥的作用。Li、Lin和Tsai(2015)以台湾省为研究案例,发现UGC在游客对目的地形象形成过程中发挥的作用主要体现在其反映的活动吸引力以及对5类主要要素(景点、购物、餐饮、住宿和交通)的情感评价上。Tseng、Wu和Morrison等(2015)分析了博客作为中国入境旅游目的地形象形成媒介所发挥的作用,并通过文本分析确定了与其相关的9个主题。Hao、Xu和Zhang(2019)基于旅游真人秀视频的“弹幕”,分析了其对Y世代观众形成冰岛认知、情感和意动形象的影响。在以上研究中,UGC均来自单一平台,Guo、Pesonen和Komppula(2021)则分析了多平台下,UGC在目的地形象投射上的差异,他们选取了5个中国本土的旅游点评平台,分析不同平台中芬兰作为目的地其投射形象的差异。另外,UGC在危机事件的传播中发挥着重要作用,进而影响游客感知形象的形成,Marine-Roig和Huertas(2020)以在巴塞罗那发生的两件公共危机事件为例,基于在线评论,分析游客对目的地形象的感知和传播,发现在线评论的分享和传播反而有助于公共危机事件下目的地形象的改善。
2.数据模态
目前用于目的地形象研究的大数据主要有文本、图片和视频3种模态,其中文本占据多数(63篇),其次是图片(14篇),视频最少(1篇)。文本在目的地形象大数据研究中的广泛应用得益于其研究方法和技术工具的成熟,用于目的地形象研究的UGC文本主要为评论文本和游记(博客),其中有一例比较特殊的是Hao、Xu和Zhang(2019)利用弹幕文本研究旅游真人秀中的弹幕对Y世代游客感知目的地形象的影响,并指出“弹幕”这种特殊文本有助于研究人员确定观众对目的地体验和感知形成的确切时刻,因而具有重要价值。
图片在目的地形象大数据研究中的应用可以分为间接分析和直接分析两种。间接分析利用图片的元数据(metadata),包含图片的标题、标签、相关描述等,分析图片含义,这样就将图片分析转化成了文本分析,如Deng和Li(2018),Deng、Liu和Dai等(2019),Fu、Wan和Wu(2022)的研究都属于这一类。直接分析则多基于深度学习技术,直接识别图片中的人物、物体和场景,获取图片含义(9篇),如Arabadzhyan、Figini和Vici (2021)使用Google Cloud Vision自动识别图片中的物体,并提取成标签,最后基于标签分析游客对目的地的认知形象;He、Deng和Li等(2022)使用DeepSentiBank(Python程序语言包)提供的深度学习工具,识别图片中的形容词—名词组合,并将其对应为目的地的情感和认知形象。
三、大数据环境下的旅游目的地形象研究:方法和技术视角
根据Echtner和Ritchie(1991)的分类,目的地形象的分析和度量方法可以分为结构化方法和非结构化方法两种。结构化方法将目的地形象先划分为不同的属性,由于这些属性一般具有共通性,便可以借助其便于编码和管理的优势,进行复杂的统计分析,也可以在不同的目的地之间进行比较,其缺点是难以度量形象的整体性和独特性;非结构化方法可以从自由形式的内容中分析目的地形象,更多采用诸如焦点小组、访谈、开放式问卷调查等研究方法,更加适于获取关于目的地的整体的、独特的形象信息,然而目的地形象信息的完整性容易受到被试者回答意愿的影响,难以开展复杂的统计分析工作,从而不易对不同目的地进行对比分析。从研究范式上来看,Echtner和Ritchie(1991)对目的地形象分析方法的分类分别对应了定量和定性两种研究范式,其中结构化分析属于定量研究,非结构化分析属于定性研究。
Lu和Stepchenkova(2015)在对UGC在旅游和接待业研究中的应用进行详细梳理之后发现,以往按照定性和定量二元分类的方式,在大数据环境下的旅游和接待业研究中已经不再适用——其中只有少量研究可以被明确判别为定性或定量研究。他们发现最为常用的研究方法是内容分析法,该方法兼具定性与定量的元素,被认为是一种混合方法。内容分析法是社会科学研究中常用的分析处理文本数据的一种非介入性方法(Babbie,2015)。通过归纳以往研究人员对内容分析法的理解,可以总结出关于该方法有定量和定性研究两种认识论:前者以文本为总体,通过定量分析做出统计推断;后者关注解释性的、非统计的方法,主要进行归纳推理(Stepchenkova,Kirilenko & Morrison,2009)。两者之间的比较如表4所示。
表 4 定量和定性文本内容分析的比较
Tab. 4
定量文本内容分析 Quantitative content analysis | 定性文本内容分析 Qualitative content analysis | |
---|---|---|
认识论 Epistemology | 实证主义(Positivism) | 解释主义(Interpretivism) |
假设 Hypotheses | 存在一个外在于人的,可以观察、度量、分析,进而理解的客观真实 | 真实不是客观的,而是一种社会和文化创造,只能被解释,但不能被完全理解 |
分析对象 Analysis object | 显性内容(Manifest content) | 隐性内容(Latent content) |
模型方法 Methodology | 列联分析、替代模型、关联模型、统计推断等 | 修辞分析、叙事分析、符号分析和话语分析等解释性的方法 |
研究立场 Research position | 代表性的(Representational) | 代表性的和工具性的(Instrumental) |
适用场景 Applicable scene | 可处理大量数据 | 难以处理大量数据 |
内容分析法是大数据环境下目的地形象研究的基本方法,尽管有些文献没有将其使用的方法(如文本挖掘、主题建模、情感分析等)称作内容分析,但正如Lu和Stepchenkova(2015)所言:“在文本(或其他符号性材料,如图片)中使用归类(categorization)和数据约减的方法,在本质上都可以划入内容分析法的范畴。”且目前大数据环境下的目的地形象研究所使用的内容分析方式以定量分析为主,定性分析多起辅助作用。定量内容分析法的认识论是实证主义,基于如下假设:存在一个外在于人的,可以观察、度量、分析,进而理解的客观真实。在此基础上,内容分析的定量化范式中的主要问题是如何进行去语境化(decontextualization)和变量的选择,中心思想是文本中的大量词语可以分类汇总形成数量较少的“类”(categories)(Stepchenkova,Kirilenko & Morrison,2009)。20世纪中期哈佛大学学者提出了列联分析(contingency analysis)方法(Stepchenkova,2012),即首先从文本中抽取不同类别,进而计算它们在文本中的共现频次,最后根据频率矩阵分析不同类别间的关系。从此,列联分析成为定量化内容分析的开端。许多学者对这一方法持有批评意见,认为它仅能够分析文本中的显性内容,不能反映文本的某些隐性特征,如修辞格、反讽等,不能将一些语境因素考虑在内,如来源的可信度,政治和社会背景,受众的年龄、性别和教育因素等,另外文本的重要性和影响力也不能仅仅通过统计数字来体现。尽管如此,定量内容分析由于其在先验设计、信度、效度、通用性和假设检验等方面的优势,在社会科学中仍得到了广泛应用。而随着计算机领域自然语言处理技术的成熟,形成了分词、去停用词、词性标注、词频统计、共现分析、关键词提取等一套自动化的文本挖掘流程,并出现了大量支持文本挖掘的软件工具(如NVivo、CATPAC II、ROST CM等)或程序语言包(如TextBlob等),为大数据环境下开展目的地形象研究奠定了技术基础。相关论文共有16篇,如Alarcón-Urbistondo、Rojas-de-Gracia和Casado-Molina(2023),Choi、Lehto和Morrison(2007),Guo、Pesonen和Komppula(2021)均是采用定量法进行文本内容分析,以达到度量目的地形象的目标。
Weber(1983)认为在内容分析中,归纳不同“类”所采用的模型方法还可以进一步分为替代模型(Substitution Model)和关联模型(Correlational Model),分别对应于机器学习领域的监督学习方法和非监督学习方法。其中,替代模型是一种监督学习方法,首先构造词典(dictionary),然后根据词典寻找文本中的同义词,将所有的同义词归为同一类,合并计算出现频率。在具体的应用中,词典的构造可以采用人工编码的方式,如Choi、Lehto和Morrison(2007),Guo、Pesonen和Komppula(2021),Iordanova和Stainton(2019),Jani和Hwang(2011),Li、He和Li等(2023),Liu、Liu和Mo等(2020),Marchi和Raschi(2022)等均是采用人工编码方式进行归类。也有许多已经开发好的知名的现成词典可供使用,比如WordNet(Fellbaum,2010)。需要指出的是,还有一类由情感词构成的情感词典(如SentiWordNet、HowNet等),使用情感词典信息对文本开展情感分析,可以获取游客对目的地的情感形象。相关论文共有12篇,如 Cao、Liang和Li等(2020),Deng和Li(2018),Fu、Wan和Wu(2022)均在研究中使用了情感词典进行情感分析。
作为一种视觉化的表达,图片和视频与文本数据一样,也蕴含了游客对目的地的认知和情感,为目的地形象分析提供了数据来源,因此目前已成为目的地形象大数据研究的热点。内容分析法同样可以用于分析视觉模态数据,虽然该方法最初被用于处理文本内容,然而正如Stepchenkova 和Zhan(2013)所指出,意象(Imagery)——无论其来自图片还是视频——也可以被当作一种文本,并利用内容分析法来做处理。目前对视觉模态内容的直接分析(即不依赖元数据)通常基于深度学习技术,主要方法有图像/视频描述(Arabadzhyan,Figini & Vici,2021)和图像识别(包含人物识别、物体识别、场景识别等)(Zhang,Chen & Li,2019;张坤、李春林、张津沂,2020),使用的工具包含深度学习框架(邓宁、蘧浪浪,2022)、程序语言包(He,Deng & Li,et al.,2022;Huang,Han & Meng,et al.,2022)和开放API(Arefieva,Egger & Yu,2021;邓宁、蘧浪浪,2022)。
四、结论和展望
本文对截至2023年4月发表在SCI、SSCI和北大中文核心期刊目录(2020版)中的与“大数据环境下目的地形象研究”主题相关的论文做了全面的收集和整理工作,并从数据、方法和技术3个视角进行了分析。在数据方面,目前用于目的地形象研究的大数据呈现多源异构的特点,数据来源以UGC为主,UGC在目的地形象研究中既可以作为分析接收形象的数据源,也可以用来对投射形象进行分析;OGC仅作为分析投射形象的数据来源,一般是用于和UGC进行对比分析;数据模态以文本数据为主,视觉模态数据(图片和视频)的应用虽然较少,但正在为学界所关注。在方法方面,目前用于大数据环境下目的地形象研究的方法从本质上来讲是内容分析法,情感分析、主题建模、图像/视频描述、图像识别等计算机科学领域方法均可以归类为此种方法。在技术方面,不同模态数据所依赖的技术有所不同,文本数据多采用基本的自然语言处理技术(包括监督学习和无监督学习),视觉模态数据多采用深度学习技术。
大数据或数据科学被称为科学研究的“第四范式”(Hey,Tansley & Tolle,2009),与基于实验、观察和理论分析的另外3个传统科研范式不同,第四范式强调数据驱动方法、先进计算技术和跨学科合作在科学研究中的变革作用,数据的指数增长以及强大的计算工具的发展使研究人员能够提高洞察力并解决复杂的科学问题。本文研究的文献可以被认为是“第四范式”在目的地形象研究领域的应用,是目的地形象研究必然要发生的范式变革。通过对数据、方法和技术三方面的梳理分析,本文认为在大数据范式下开展目的地形象研究的特点和优势主要有:(1)无论是横截面分析还是时间序列分析,相关研究多以研究范围内的全量UGC为数据来源,进行目的地形象分析和度量,避免了抽样方法差异对结果造成的影响,研究结论具有较好的泛化性;(2)与其他研究方法相比,大数据范式下的研究方法能够自动地、批量地分析处理散布在文本、图片、视频中的大量非结构化目的地形象数据,更为高效;(3)如前文所述,大数据范式下的目的地形象研究所采用的研究方法均可归类为内容分析法,数据来源为二手数据,采用的是非介入的研究方式,避免了对研究对象的干扰,通过UGC可以了解到游客对目的地形象较为真实的感知。
目前大数据范式在目的地形象研究中的应用也有其不足之处,主要表现为:(1)目前应用于目的地形象研究中的大数据技术,如主题建模、情感分析、图片描述等,仅限于量化分析文本、图片和视频中的显性内容,而其中的隐性内容仍需要采用符号分析等人工方式进行;(2)与其他研究范式不同,大数据分析通常可以揭示数据之间的相关关系,即某些变量之间的关联或趋势,然而却难以确定因果关系,应用于目的地形象研究时同样如此;(3)受制于海量数据的一致性、标准化、数据质量以及机器学习算法准确性的影响,与人工分析相比,大数据范式下对目的地形象相关概念的测量精度较低。本文认为大数据分析的优缺点既由其“数据驱动”的范式特点所决定,同时也受到数据、方法和技术等方面的影响。因此,为了避免大数据研究范式的不足,在目的地形象研究中,一是既要尽量发挥出大数据分析的优势,又要注重跨学科的融合,将“大”数据与“小”数据环境下的定量、定性方法相结合,共同应用于目的地形象分析;二是要做好数据的清洗和预处理,提升用于目的地形象分析的数据质量;三是持续跟踪大数据领域的前沿技术,并将之用于目的地形象分析,提升分析的精准性。
通过对目前大数据环境下目的地形象研究的总结和分析,并结合当前计算机和信息科学领域相关技术前沿状况和未来发展趋势,对于未来大数据环境下的目的地形象研究,本文认为应重点关注3个“新”,即新数据来源(人工智能生成内容)、新数据模态(视觉模态)、新技术方法(多模态学习),具体如表5所示。
表 5 未来可能的研究方向
Tab. 5
大数据层面 Level of big data | 未来可能的研究方向 Possible future research direction |
---|---|
新数据来源 (人工智能生成内容) | 将AIGC作为目的地投射形象的数据来源,研究如何提升目的地形象投射的精准性,提高目的地投射形象和游客感知形象的一致程度 |
将AIGC作为游客感知形象的数据来源,研究如何基于AIGC度量游客感知形象,分析目的地形象的形成过程 | |
新数据模态 (视觉模态) | 对图片和视频内容进行直接分析,度量与目的地相关的游客感知形象 |
挖掘图片和视频中与游客感知形象相关的隐性内容 | |
新技术方法 (多模态学习) | 利用多模态数据融合和跨模态数据生成(如基于图片生成文本)技术提升目的地形象度量的准确性 |
(1)在数据来源上,未来应该重点关注AIGC在目的地形象研究中的应用。AIGC是继OGC和UGC之后,利用人工智能进行的一种新的内容生产创作方式和技术手段。随着以“生成式对抗网络”为代表的深度学习技术的提出,AIGC进入了快速发展阶段(中国信息通信研究院和京东探索研究院,2022),尤其是近期随着GPT-3.5、GPT-4等大型语言模型的出现,ChatGPT一跃成为热门的消费级应用程序,AIGC开始进入公众视野,也日益为学术界所关注,成为当前的研究热点。AIGC在内容生产上的变革性对于许多学科都有外溢效应,具体到目的地形象研究方面,AIGC在旅游市场的供给侧和需求侧均可以发挥重要作用,影响目的地形象的投射和接收。在供给侧,借助AIGC创作的灵活性和自动化特点,DMO可以进一步降低营销成本,同时可以根据游客特点和需求,精准化、个性化地生成相应的目的地形象,有效提高游客对目的地形象的接受程度;在需求侧,AIGC的出现赋予了游客除拍照、录像之外更为丰富、生动的表达目的地形象感知的方式。通过AIGC,研究人员可以更为准确全面地度量游客对目的地形象的认知和情感,更为细致地分析游客感知形象的形成过程。
(2)在数据模态上,应关注视觉模态数据在目的地形象研究中的应用。目前大数据环境下的目的地形象研究主要集中于文本模态,较少使用图片、视频数据,且目前部分尝试使用图片数据的研究也多是采用间接分析(基于元数据)或人工编码的方式,未能真正发挥图片、视频的大数据优势,因此未来研究应聚焦于图片视频内容的直接分析,丰富并完善目的地形象的度量方法。另外,在当前基于大数据的目的地形象研究中,对图片含义的自动化分析还停留在显性内容层面,缺乏对图片隐性内容的挖掘。随着大型语言模型的发展,图片隐含含义,诸如讽刺、幽默、比喻、夸张等,也能够被自动解析,因此后续目的地形象研究在提取图片和视频显性内容的同时,还应关注对隐性内容的挖掘和建模,这必将有效提高游客感知形象度量的准确性。
(3)在方法和技术上,应重视多模态学习(multimodal learning)在目的地形象研究中的应用。多模态学习是多模态机器学习的简称,指通过建模从多种模态数据中进行学习,实现不同模态数据之间的融合、协同和转化。多模态深度学习是多模态机器学习发展到现阶段的必然产物(刘建伟、丁熙浩、罗雄麟,2020)。整体来说,计算机领域关于多模态学习理论、模型和算法的研究要远快于其在目的地形象研究领域的应用。目前在目的地形象研究中,深度学习模型和技术尚未全面应用,主要集中于为数较少的基于图片和视频数据的目的地形象研究,文本数据的分析尚停留在基本的自然语言处理和主题建模阶段,较少使用深度学习方法,更没有实现多种模态数据融合和相互转化的目的地形象研究案例。展望未来多模态学习在目的地形象研究中的应用,应聚焦于发挥多种模态数据之间的协同作用,这既能为目的地形象研究提供更加丰富的方法和技术支撑,又能通过多种模态数据的协同使用提高研究结果的信度、效度和泛化性。另外,还应关注跨模态数据生成技术在目的地形象研究中的使用,如基于图片生成摘要文本的图片描述技术、图片情感识别技术等,它可以有效识别图片等视觉模态数据内的隐性内容,提升目的地形象度量的准确性。
本文将对大数据环境下目的地形象研究现状的分析和对未来研究的展望,经过归纳整理和逻辑推导,总结为图2所示的研究框架。该框架由两个维度、4部分内容所组成。从横向维度看,左侧列出的是与目的地形象研究相关的大数据要素,主要从数据、方法和技术(工具)3个层面进行分析,右侧列出的是在大数据的赋能下与目的地形象相关的研究方向。从纵向维度看,上方列出的大数据环境下目的地形象研究的现状,对应了本文第二、三两部分的综述分析,下方列出的是未来在大数据环境下与目的地形象相关的可能开展的研究方向。该框架能够帮助相关领域的研究学者了解目的地形象研究现状,为其探究该领域未来研究方向提供路径参考,并对相关从业者开展旅游营销实践提供方法论上的指导。
图 2
图 2
大数据环境下目的地形象研究框架
Fig. 2
Destination image research framework under the big data environment
参考文献
Proposal for employing user-generated content as a data source for measuring tourism destination image
[J].
Exploring destination’s negative e-reputation using aspect based sentiment analysis approach:Case of Marrakech destination on TripAdvisor
[J].
Measuring destination image:a novel approach based on visual data mining. A methodological proposal and an application to European islands
[J].
A machine learning approach to cluster destination image on Instagram
[J].
A model of destination image formation
[J].
Latent dirichlet allocation
[J].
TDIVis:Visual analysis of tourism destination images
[J].
Destination image representation on the web:Content analysis of Macau travel related websites
[J].
Feeling a destination through the “right” photos:A machine learning model for DMOs’ photo selection
[J].
Different cultures,different photos:A comparison of Shanghai’s pictorial destination image between East and West
[J].
The meaning and measurement of destination image
[J].
Inbound tourists’ perception of tourist destination image classified by UGC picture computer program
[J].
Destination image:Towards a conceptual framework
[J].
Image formation process
[J].
Comparing online travel review platforms as destination image information agents
[J].
Barrage participation and feedback in travel reality shows:The effects of media on destination image among Generation Y
[J].
How to “Read” a destination from images? Machine learning and network methods for DMOs’ image projection and photo evaluation
[J].Online photos can reflect tourists’ received destination image and be used to project destination image by destination marketing organizations (DMOs). Studies have identified a gap between projected and received images, highlighting the difficulty DMOs face when selecting content to project the “right” image. Taking an audience-driven perspective, this study analyzed information from user-generated content (UGC) to guide the selection of organization-generated content (OGC) on social media. Using a machine learning algorithm, we extracted connected cognitive and affective elements of received and projected images from UGC and OGC. The elements and their relationships retrieved from UGC were then used to construct a semantic network. The network informs the core–periphery structural information of each element and guides DMOs’ image projection and content selection. Studies with two independent samples demonstrated that an OGC photo whose projected images matched consumers’ central impressions, particularly affective ones, could induce higher online engagement.
Destination image recognition and emotion analysis:Evidence from user-generated content of online travel communities
[J].The tourism destination image is an intangible value that enhances the internal and external spiritual value of the region. To improve tourist experiences and provide reference for relevant departments, we applied the GooSeeker web data crawler tool and Python data mining kit to crawl and analyze the representative online tourism community data. We conduct an empirical analysis through data from the online tourist community ‘mafengwo’. The result, based on the user-generated content data analysis of online travel community, shows that the tourists' perception of the destination image, cognitive theme and emotional experience has different effects on the tourist experience. This research offers insights into destination image cognitive theme and traveler behavior habits, which can provide guidance for platform and destination managers.
Do the DMO and the tourists deliver the similar image? Research on representation of the health destination image based on UGC and the theory of discourse power:A case study of bama,China
[J].Even though destination image is an important expression of discovering the local landscapes and place significance, the construction and measurement of destination image neglect the place component. This research explores the image of health destinations, as well as its representation mechanism, combining the triadic structure of tourism image proposed by Marine-Roig et al. with the theory of discourse power put forward by Michel Foucault, taking Bama, Guangxi as a case. In addition, this paper uses the IPA matrix to visually unveil the pronounced gap between the projected image by Destination Management Organizations (DMOs) and the perceived image of tourists and suggests strategies that DMOs should adopt in the different dimensions.
Image as a factor in tourism development
[J].
Cognition, emotion and trust:A comparative analysis of Cambodia’s perceived and projected online image
[J].A comparative analysis of information on Cambodia as a tourist destination published on official tourism websites and amateur travel blogs suggests that there are substantial differences in the way Cambodia's destination marketers and commercial enterprises promote the destination and the way travel bloggers perceive it and write about it. Cambodia's projected online image elicited through the analysis of official tourism websites consists of mainly cognitive image elements (knowledge and beliefs about Cambodia), whereas its perceived image obtained from amateur travel blogs contains both cognitive and affective (feelings and attitudes towards Cambodia) image elements. Furthermore, the 'friend'-style relationship that the reader develops with the travel blogger can be attributed to personal projections and thus induces an element of trust not established between the reader and the websites. Findings suggest that a lack of perceived trust among destination marketers and an increased level of trust among authors of user-generated content (UGC) and their readers has led to destination marketers being at a distinct disadvantage compared with their blogging colleagues. The use of affective evaluations of destinations by bloggers alongside their perceived credibility is likely to weigh more heavily in consumer's decision-making process than promotional material presented by the likes of tourist boards and commercial enterprises. The theoretical and practical implications of these findings are also discussed.
User-generated destination image through weblogs:A comparison of pre- and post-visit images
[J].
Analyzing destination branding and image from online sources:A web content mining approach
[J].
Tourism destination image:Conceptual problems and definitional solutions
[J].Although tourism destination image (TDI) has been extensively studied, the nature and scope of TDI remain vague. This study aims to address this conceptual problem from a modernist perspective. Forty-five representative TDI definitions are analyzed, and a new definition is proposed by adopting a “seven-step” procedure derived from definition theory in logic studies. Results show that (1) currently TDI is defined mainly as the mental/total impressions/perceptions held by tourists pertaining to a destination, (2) such definitions are quasi-theoretical in type and created by roughly following the connotative definition technique, (3) the quality of these definitions is generally low, and (4) the new definition proposed in this study better captures the essence of TDI and considerably reduces TDI’s internal and external vagueness. In defining a key tourism concept, this study has notable implications for advancing TDI research and defining tourism concepts more rigorously.
Core-periphery structure of destination image:Concept,evidence and implication
[J].
A cross-cultural anatomy of destination image:An application of mixed-methods of UGC and survey
[J].
Keep it real:Assessing destination image congruence and its impact on tourist experience evaluations
[J].
Impact of short food videos on the tourist destination image-take Chengdu as an example
[J].Taking Chengdu as an example, based on the destination image theory and employing the content analysis methodology, this paper conducts data mining on the online comment texts of TikTok short food videos, and analyzes the impact of short food videos on the destination image (cognitive image, affective image and conative image). The results show that: (1) in terms of cognitive image, short food videos have increased potential tourists’ attention to the destination image, especially their attention to the flavor characteristics of food in the destination and the local social environment; (2) in terms of affective image, the comments of short food videos are mainly neutral and positive, and the contents about the flavor characteristics of food and the local social environment are more likely to affect the affective image of the destination; and (3) in terms of conative image, the appearance description of food in short food videos brings about an obvious effect of intention, and it also creates the demand to travel together and obtain information. This paper is inspiring for city managers and tourism marketers to use TikTok short videos to establish and disseminate food-based city brands and destination images.
Traveller-generated contents for destination image formation:Mainland China travellers to Taiwan as a case study
[J].
Using text mining to track changes in travel destination image:The case of Macau
[J].
Using big data in the academic environment
[J].
User-generated content as a research mode in tourism and hospitality applications:Topics,methods,and software
[J].
Exploring the destination image based on the perspective of tourists’ expression using machine learning methods combined with PLTS-PT
[J].
Exploring destination image through online reviews:An augmented mining model using latent Dirichlet allocation combined with probabilistic hesitant fuzzy algorithm
[J].In the era of information overload, the density of tourism information and the increasingly sophisticated information needs of consumers have created information confusion for tourists and scenic-area managers. The study aims to help scenic-area managers determine the strengths and weaknesses in the development process of scenic areas and to solve the practical problem of tourists' difficulty in quickly and accurately obtaining the destination image of a scenic area and finding a scenic area that meets their needs.
Measuring destination image of an Italian island:An analysis of online content generated by local operators and tourists
[J].
Measuring destination image through travel reviews in search engines
[J].
Perceived image specialisation in multiscalar tourism destinations
[J].
Measuring the gap between projected and perceived destination images of Catalonia using compositional analysis
[J].
How safety affects destination image projected through online travel reviews
[J].
Regional images and regional travel behavior
[C]//
The long tail of destination image and online marketing
[J].
Destination image analysis—A review of 142 papers from 1973 to 2000
[J].
Big data:A review
[C]//
Facilitating content analysis in tourism research
[J].The article proposes a new methodological approach to facilitate content analysis of electronic textual data in a more efficient and transparent way. The textual data are processed iteratively by two software products, CATPAC and WORDER. This approach permits smoothing of the original textual data, identification of the variables of interest, frequency count of the occurrences of these variables in the texts being processed, storage of frequency results in general purpose statistical packages, and subsequent dimensional reduction of word-frequency data by means of factor analysis. Application of this methodology is illustrated on three examples of destination-image studies, which cover content analyses of open-ended responses to e-survey questions, texts from tourism Web sites, and newspaper articles. Advantages and disadvantages of the proposed research technique, its contribution to tourism studies, and the place of the approach within the quantitative paradigm of content analysis are also discussed.
Chinese outbound tourists’ destination image of America:Part II
[J].This study compares the U.S. destination image of four groups of Chinese travelers with different travel horizons. The authors attempt to explore new and effective ways of comparing qualitative image information using methodological approaches from other disciplines. Specifically, they introduce and examine the concepts of destination image richness, evenness, and dominance, analogous to such concepts in biology as species richness, species evenness, and species dominance. Furthermore, the pattern of destination image distribution for the four groups is also explored from a perspective of a power law distribution often used in economic analyses. The obtained results partially support the travel horizon proposition. By applying ideas from other disciplines to the study of destination image, the authors hope that this study contributes both to the image literature and interdisciplinary research on tourism in general.
Visual destination images of Peru:Comparative content analysis of DMO and user-generated photography
[J].
Travel blogs on China as a destination image formation agent:A qualitative analysis using Leximancer
[J].
Tourism destination image based on tourism user generated content on internet
[J].The purpose of this paper is to study tourists’ spatial and psychological involvement reflected through tourism destination image (TDI), TDI is divided into on-site and after-trip groups and the two groups are compared in the frame of three-dimensional continuums.
Measurement models for content analysis
[J].
Discovering the tourists’ behaviors and perceptions in a tourism destination by analyzing photos’ visual content with a computer deep learning model:The case of Beijing
[J].
Reconsidering tourism destination images by exploring similarities between travelogue texts and photographs
[J].
基于UGC文本数据分析的黑色旅游目的地形象感知研究
[J].
On image perception of dark tourism destination based on UGC text data analysis
[J].
基于文本挖掘的景区旅游形象感知研究——以杭州西溪国家湿地公园为例
[J].
Research on tourism image perception of scenic spot based on text mining:A case study of Hangzhou Xixi national wetland park
[J].
基于网络文本分析的重游意向旅游形象感知研究——以厦门鼓浪屿风景名胜区及北京故宫博物院为例
[J].
Research on tourist perception of revisit intentions based on network text analysis:Taking Gulangyu scenic spot in Xiamen and Beijing palace museum as examples
[J].
冬奥会申办成功对北京旅游目的地感知形象的影响
[J].
Impact of successful bid for winter olympics on perceived image in Beijing tourism destination
[J].
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
[J].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。
Different perceptions of Beijing’s destination images from tourists:An analysis of Flickr photos based on deep learning method
[J].
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.
基于视频机器分析的目的地形象差异对比——以北京YouTube视频为例
[J].
Comparison of destination images based on video analysis through machine learning:A case study on YouTube videos of Beijing
[J].
基于UGC图片元数据的目的地形象感知——以北京为例
[J].
Perception of travel destination image based on user-generated photograph metadata:The case of Beijing
[J].
基于网络游记的陕西旅游目的地形象感知研究——以陕西省八大5A级旅游景区为例
[J].
Research on image perception of Shaanxi tourism destination based on network travel notes analysis:Taking eight national 5A scenic spots in Shaanxi Province as an example
[J].
入境游客对颐和园的旅游目的地形象感知——基于语料库的研究
[J].
Exploring inbound tourists’ perceptions on the tourism destination image of the Summer Palace:A corpus-based study
[J].
基于网络评论的五台山旅游目的地游后形象感知研究
[J].
Post-visit image perception of Mount Wutai tourism destination based on online comments
[J].
多模态深度学习综述
[J].
Survey of multimodal deep learning
[J].
大数据研究综述
[J].
Survey of big data
[J].
基于网络大数据的上海红色旅游形象感知与情感评价研究
[J].
A research on image perception and emotional evaluation of Shanghai red tourism based on network big data
[J].
2022年冬奥会对张家口城市旅游地形象的影响研究——基于UGC文本分析
[J].重大节事活动对旅游地形象的影响具有时序动态特征,分析其内在机理有利于促进目的地营销和竞争力提升。本文采用Python数据挖掘技术和自然语言处理技术(NLP),爬取国内知名旅游门户网站的旅游评论,从“认知-情感-整体”3个维度构建分析框架,探究2022年冬奥会申办期、筹备期、预热举办期的张家口城市旅游地形象的时序变化特征,解析特殊时期冬奥会对城市旅游地形象的潜在影响和作用机制。结果表明:① 张家口市的城市旅游地形象构成要素日趋多元化,冬奥会的作用效果由弱变强。张家口市的冰雪旅游形象关键度不断提升,经历了由模糊向具象转变的细化过程,情感形象产生积极变化;② 冬奥会对张家口城市旅游地形象产生了渐进式影响效应,通过旅游体验的建构作用、投射形象的引致作用和品牌感知的晕轮作用发挥效果。冬奥会推动张家口市的产品业态创新、基础设施升级及服务水平提升,通过改善游客体验对城市旅游地形象产生积极影响。研究结果对东道主城市的旅游营销和规划具有一定指导意义。
Research on the influence of 2022 Winter Olympic Games on the tourism destination image of Zhangjiakou:Based on UGC text analysis
[J].
The impact of mega-events on the tourism destinations image (TDI) has a time-series dynamic feature, and the analysis of the inner mechanism is conducive to promoting destination marketing and competitiveness. This paper adopted Python data mining technology and natural language processing technology (NLP) to crawl the travel reviews of well-known domestic travel websites and established a dataset based on long time series and large sample data. The analysis framework is constructed from the “cognitive-emotional-overall” dimensions to explore the temporal change of Zhangjiakou's TDI during the bidding period, preparation period and warm-up period of the 2022 Winter Olympic Games. In this way, the potential impact and mechanism of the Winter Olympics on the TDI in a special period and in a specific way can be analyzed. The results show that: (1) The composition elements of Zhangjiakou's TDI are becoming increasingly diversified and the effect of the Winter Olympics is becoming more and more significant. The key degree of ice and snow tourism image under the influence of the Winter Olympics is increasing, and it has experienced the refinement process of changing from vague to concrete image. Positive emotions are becoming more prevalent. (2) The Winter Olympic Games have a progressive influence on the TDI of Zhangjiakou City, which exerts its effect through the construction of tourism experience, the projection of image and the halo effect of brand perception. The Winter Olympics promote product innovation, infrastructure upgrade and service level improvement in Zhangjiakou, which will have a positive impact on the TDI by improving the visitor experience. This paper explores the dynamic impact path of the Games on the changing image of city tourism, which is a guide for host cities to change the TDI through festivals and post-event marketing.
基于图片大数据的入境游客感知和行为演变研究——以北京市为例
[J].
A study on the spatio-temporal evolution of inbound tourists’ perceptions and behaviors based on big pictorial data:The case of Beijing
[J].
基于网络文本的目的地旅游形象游客感知与官方传播对比研究——以福建永定土楼为例
[J].
A comparative research on tourism destination image of tourists’ perception and official dissemination based on network text:Taking Yongding Tulou in Fujian as the case
[J].
旅游形象研究中问卷调查和网络文本数据的对比——以西安旅游形象感知研究为例
[J].
A comparative study on the data of a questionnaire and web texts in tourism image researches:A case study of Xi’an’s tourism image perception
[J].
人工智能生成内容(AIGC)白皮书
[R/OL].
White Paper on Artificial Intelligence Generated Content (AIGC)
[R/OL].
/
〈 |
|
〉 |
