Titre : Multidimensional analysis of social network content: strategies, calculation and visualisation of indicators for tourism 

Auteurs : Maxime Masson
, Rodrigo Agerri
, Christian Sallaberry
, Marie-Noelle Bessagnet
, Philippe Roose
, Annig Le Parc Lacayrelle,  

Revue : Open Journal in Information Systems Engineering 

Numéro : Issue 1 

Volume : 5 

Date : 2025/06/4 

DOI : 10.21494/ISTE.OP.2025.1303 

ISSN : 2634-1468 

Résumé : The growing influence of social networks in the field of tourism highlights the need for effective automatic natural language processing (NLP) approaches to exploit this resource. However, transforming multilingual, informal and unstructured texts into structured knowledge remains a challenge, not least because of the need for annotated data for model training. This paper first examines different learning-based NLP techniques and models to optimise performance while reducing the need for manually annotated data. A new multilingual dataset (French, English, Spanish) specific to tourism has been created, focusing on the Basque Country region. This dataset includes tweets with manual annotations on spatial named entities, tourism thematic concepts and sentiments. A comparison of fine-tuning and few-shot learning methods with multilingual models indicates that few-shot techniques can produce good results with few annotated examples. Experiments conducted on this dataset suggest the possibility of applying learning-based NLP methods to various domains, while reducing the need for manual annotations and avoiding the complexities of rule-based methods. Secondly, we present TextBI, a generic multimodal dashboard designed to present multidimensional text annotation analysis on large volumes of multilingual social media data. The tool focuses on several dimensions: spatial, temporal, thematic and personal, and also supports additional enrichment data such as sentiment and engagement. Our approach, TextBI, represents a significant advance in the field of NLP annotation results visualisation by integrating and blending features from a variety of Business Intelligence, Geographic Information Systems and NLP tools. 

Éditeur : ISTE OpenScience