Table 3.

Methods used in AI-driven hospitality personalization

DocumentSample areaPeriod of studySample sizeSampling techniquesMethods/techniques
Lin et al. (2025) Study 1Instagram posts tagged with Saguaro National ParkCollected March 2022– March 202314,906 Preprocessed using LIWC-22 for keyword frequency and tone scoring
Study 2U.S. adults (18+ years) who visited a natural area in the past 12 months2023400 surveyed → 357 valid responsesProlific (using demographic balancing)Fuzzy set qualitative comparative analysis (fsQCA)
Chen et al. (2025) Social media posts from Xiaohongshu and Dianping, using hashtags like #HongKongTravel, #HongKongTourFebruary 28, 2018, to October 18, 202310,261UGC (user-generated content) collected via octopus collector and python scrapingPython (octopus collector, jieba), PaddlePaddle (deep learning), BERT embeddings, K-means, TF-IDF, dynamic McKinsey matrix modeling
Mahmoud et al. (2024) Social media comments scraped (Reddit, TikTok, YouTube) → filtered down to 2,650 high-quality, English-language, nonduplicate commentsJanuary- 2020-December 202215,461Big-data, nonprobabilisticPython (scraping, VADER sentiment), NVivo (qualitative coding), matplotlib (visuals)
Pitakaso et al. (2025) Thailand: case study in Ubon Ratchathani ProvinceNA316 hotel guests, 16 chambermaids, five managersCase study-based purposive sampling in a real-world hotel environmentAI framework: Artificial multiple intelligence system (AMIS) combining GA, DE, PSO, ACO, and VaNSAS to solve a multi-objective scheduling problem
Mensah et al. (2025) 10 most popular African destinations (e.g. Nairobi, Marrakesh, Cape Town, Arusha)August 2018 – August 20235,043 Tripadvisor reviewsPurposeful selection of top-ranked destinations based on Tripadvisor Travelers’ Choice (2023)Quantitative sentiment analysis using python, TextBlob, pandas, and seaborn. Keyword frequency analysis. Content analysis using tourist reviews for theme generation
Zhang et al. (2025) TripAdvisor, Houston, Texas marketSep-2332,129 hotel reviews with managerial responsesPurposive sampling (only reviews with hotel staff responses were selectedGPT-3.5 response generation (Temp 0,1,2), BERT-based cosine similarity, t-SNE visualization, Negative Binomial Regression
Tan et al. (2025) Chengdu, China (6 districts)January 2014 – January 202323,138 valid reviews from 243 storesRandom selection of stores; exclusion of manipulated/incomplete reviewsWeb scraping, sentiment analysis (dictionary + bayesian model), fsQCA (fuzzy-set qualitative comparative analysis), machine learning
Yi et al. (2024) AI-themed hotels in China (12 hotels across seven brands)Jun-227,431 online customer reviewsPurposive sampling (based on hotel names and descriptions explicitly referencing AI)Topic modeling (LDA), fuzzy-set qualitative comparative analysis (fsQCA)
Guidotti et al. (2025) European castles, U.S. hotels, global restaurants (from Google reviews)Not explicitly mentioned (datasets are pre-existing from Kaggle)13,650 reviews (2,550 castles, 1,100 restaurants, 10,000 hotels)Purposive sampling (publicly available datasets selected for tourism context)Zero-shot learning with LLMs (DeBERTa, RoBERTa, BART); Sentiment analysis; Keyword extraction; NLI-based prompting; Statistical analysis (accuracy, precision, recall, F1, IQR, median)
Félix et al. (2025) TripAdvisor reviews from users worldwide (seeded from 5 cities: Tiradentes, Ouro Preto, San Gimignano, Cannes, Ibiza)October 2001 to May 2020420,987 labeled English reviews (from 65,812 users, 29,176 cities)Snowball sampling starting from seed citiesText classification using ML (Logistic Regression, LGBM), Transformers (BERT, RoBERTa, BART), LLMs (LLama2, Bloom); TF-IDF, contextual embeddings, domain-specific feature extraction
Kirilenko and Stepchenkova (2025) TripAdvisor (arenal and corcovado), YouTube (Li Ziqi videos), Weibo (Li Ziqi posts)2011–2022∼26,000+ user-generated documents (6 datasets)Purposive sampling of tourism-related UGC from 3 platformsTopic modeling using LDA (sklearn), BERTopic (MiniLM), and GPT-4 (OpenAI API); Manual validation and comparative analysis across five performance dimensions
Anguera-Torrell and Nicolau (2025) Barcelona, Spain2018–202222,326 reservationsPurposive sampling (hotel bookings via booking.com)Linear probability model (LPM), logit and probit models, interaction analysis, fixed effects regression
Chotisarn and Phuthong (2025) Western Thailand’s coastal tourism destinationsAugust–September 2024186Purposive samplingPLS-SEM (SmartPLS 4), Descriptive Statistics (SPSS 27), Harman’s single-factor test
Kourtit et al. (2025) LondonPre-COVID and during-COVID458,078 Airbnb + 20,417 hotel reviewsListings with ≥10 reviews per periodText Analysis, Sentiment Analysis (sentimentr in R), Multilevel Analysis (3-level), Geographically Weighted Regression (GWR)
Piriyakul et al. (2024) Amari Phuket hotel, ThailandYear 2022250 customer reviews (randomly selected from TripAdvisor)Random sampling from publicly available review dataText mining, sentiment analysis, K-Means clustering, naive bayes, SEM (structural equation modeling)
Yu et al. (2024) Sichuan province, ChinaReviews collected from 2012 to 2023 across platforms1,003,501 valid reviews from Meituan + 1,596,398 texts from Sogou News, Wikipedia, CtripPurposive filtering of hospitality-related content and de-duplicationDictionary-based Sentiment Analysis, Word2Vec, HowNet Similarity, Manual Validation
Li et al. (2024) 5 cities in China (Beijing, Shanghai, Hangzhou, Shenzhen, Chengdu)Not explicitly mentioned (data from Dianping.com up to 2023)67,060 reviews collected; 1,994 reviews manually analyzedPurposeful selection of spacelab restaurant chainsText mining (ROST CM), Co-word clustering (gephi), sentiment analysis (SnowNLP), logistic regression
Blanco-Moreno et al. (2024) Camino de Santiago (UNESCO WHS)2010–202227,088 postsWeb scraping (targeted sampling)AI-based content analysis (ML for text, DL for images), generalized linear model (UGLM)
Camatti et al. (2025) Venice, Italy (15 fortress sites)Early 202415 fortresses; 2 expert groups (GE1 and GE2); 40 experts in GE2Purposive sampling of experts (academics, planners, community stakeholders)Fuzzy expert systems (FES), Self-Organizing maps (SOM), MCDA
Firza et al. (2024) Italian provinces2004–2020107 provincesCensusCanonical correlation analysis (CCA), fuzzy C-Means clustering (FCM), random Forest classification (supervised ML), SHAP (XAI)
Gupta and Jaiswal (2024) Online hotel review platformsNot explicitly mentioned23,806 reviewsWeb scrapingTextBlob and VADER LDA, GSDMM (gibbs sampling dirichlet mixture model), logistic regression, naïve bayes, random Forest, AdaBoost, SVM, neural networks
Source(s): Developed by authors
[ViewLarge]
This Feature Is Available To Subscribers Only

or Create an Account

Close subscription notice
Close access options