Semantic Diversity and Visitor Sentiments in Ecotourism Using Geospatial Natural Language Processing of Social Sensing Data

This study aimed to investigate visitors’ perceptions of ecotourism landscapes in Thailand’s protected areas by integrating spatial, semantic, and affective information derived from social-sensing textual data. To this end, it examined the influence of ecosystem characteristics on thematic expressions and emotional tones across five national parks. Methodologically, a spatio-semantic, natural language processing (NLP) social-sensing framework was developed using geotagged Flickr tags and YouTube comments from international (English) and domestic (Thai) tourists. Textual data were preprocessed (cleaned, normalized, and tokenized) and analyzed using latent Dirichlet allocation (LDA), sentiment analysis, and diversity metrics. Concurrently, YouTube comments were similarly processed using LDA and rule-based sentiment analysis. Subsequently, a Diversity × Topic × Season matrix integrated Flickr-derived indicators with YouTube-derived sentiment and topic dominance. Binary logistic regression was applied to examine cross-platform relationships. The analysis identified three dominant themes, namely Nature & Landscapes, Travel & Activities, and Feelings & Experiences. Forest-mountain parks showed high semantic diversity and strong positive sentiment, whereas marine parks exhibited narrower but predominantly positive activity-driven discourses. Moreover, seasonal variation was evident, with summer and winter yielding the highest diversity and positivity. Building on these results, this study devised a spatio-semantic framework for analyzing variation in visitor expressions across parks and seasons. It captures eco-awareness, perceptions, and preferred activities in protected landscapes. From a practical perspective, semantic diversity and sentiment indicators can support ecotourism management through improved visitor monitoring and communication strategies.