Finding a suitable POI based on the user's needs and intentions is a complex decision-making process. Obtaining valuable information from the vast amount of social media data and using it for travel recommendations is a challenging issue. Traditional POI recommendation algorithms do not fully take into account the true feelings of customers about tourist attractions implied in social media data because they usually require a large amount of tagged travel commentary data. This study presents an aspect-based sentiment analysis model and POI recommendation method to accurately capture sentiment information contained in social media data with a small amount of tagged data. The pre-training model BERT is used to obtain the embedded representation of words that fuse the semantic information of the text. Using contrastive learning, point clusters belonging to the same class in the embedded space of words are pulled together, and sample clusters from different classes are separated. The potential relationship between comment ratings and their impact on user perception is analyzed, and the best performance formula for the loss function is determined. The test accuracy and F1-Score of the model in the experiment improved by 13.03% and 12.23%, respectively, compared to the BERT base model. POI recommendation validation is performed using a variety of recommendation algorithms. The experimental results show that the addition of aspect-based sentiment attributes can effectively improve the accuracy of recommendations.
Keywords: BERT; Contrastive learning; POI recommendation; Sentiment analysis.
© The Author(s), under exclusive licence to Springer-Verlag GmbH Austria, part of Springer Nature 2023, Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.