Auxiliary Information Enhanced Multi-Task Recommendation for Tourist Attractions

Attraction recommendation systems can help tourists filter irrelevant information, improve the accuracy of recommendations and tourists' satisfaction, and explore potential business opportunities for merchants. However, the existing attraction recommendation systems focus on utilizing the interaction history information between tourists and attractions and need more auxiliary information about attractions, which limits the accuracy of recommendations. This paper presents a multi-task attraction recommendation model that utilizes auxiliary information enhancement. The model utilizes attraction attribute information to construct a knowledge graph and employs a pre-trained BERT model as an encoder to establish a multi-task augmented attraction auxiliary information representation learning framework. Ultimately, it predicts by taking the dot product of user and attraction features. The model's performance has been evaluated on real-world data of the “One Mobile Phone Touring Yunnan” application, and it achieves notable results in precision, recall, and NDCG.