Developing a web application for co-living

Analysis of User-Generated Content in Visitor Reviews of Tourist Attractions Using Semantic Similarity

Online tourist reviews, a major form of user-generated content (UGC), are often short and unstructured, complicating the identification of tourist experience dimensions and their relationships. This study presents an integrated framework combining Topic-RoBERTa, topic-level semantic network analysis, and Graph Attention Networks (GAT) to extract experience topics and model their semantic associations. The authors validated the framework on 186,429 reviews from the OD-TripM TripAdvisor review dataset released by the Data Science and Computational Intelligence (DaSCI) research group on GitHub and the Yelp Open Dataset, identifying 12 tourist experience dimensions and constructing a community-structured semantic network. The results show that the proposed method can extract interpretable topics, reveal fine-grained attention-weighted associations, and provide a more structured understanding of tourist experiences. The findings offer valuable insights for tourism management and service optimization.

Restaurant Recommendation Algorithm Based on Category Preferences and User Interests

With the rapid development of the Internet and local life services, catering information has grown explosively, making it difficult for users to quickly find dining venues matching their preferences. To address information overload, homogeneous recommendations, and insufficient semantic understanding, this paper proposes a restaurant recommendation algorithm based on lightweight three-dimensional fine-grained semantic quantification and user interests. First, catering review texts are preprocessed, and a text representation model combining Word2Vec (word embedding) embeddings and TF-IDF (term frequency–inverse document frequency) weighting is constructed. An SVM (support vector machine) classifier is then used to categorize reviews into taste, environment, and service, enabling quantification and structuring of unstructured data. Next, user preference vectors and restaurant feature vectors are built from classification results. A hybrid strategy integrating cosine similarity and category label fitness is designed to generate a Top-5 personalized recommendation list. Experiments on 23,998 real reviews of Nanxin Dessert Shops show the SVM classifier achieves 90.7% overall accuracy, and the average comprehensive matching degree of recommendations is close to 0.9. The proposed method effectively mines users' fine-grained preferences, improves recommendation accuracy and personalization, and provides a feasible solution for catering platform recommendation systems.

for City Exploration

An Empirical Evaluation of Cross-City POI Recommendation on a Large-Scale Benchmark

The need for task-specific designs that support cross-city preference transfer, semantic grounding, and scalable reasoning over unseen destination inventories is highlighted, with results highlighting the need for task-specific designs that support cross-city preference transfer, semantic grounding, and scalable reasoning over unseen destination inventories.

An AI-Powered Location-Based Recommendation System for Personalized Urban Mobility

The increasing availability of heterogeneous spatial and contextual data sources presents both an opportunity and a challenge for location-based decision support systems. This paper presents an AI-powered recommendation system designed to facilitate personalized urban mobility while enhancing users' sense of place. The proposed system integrates multiple open data sources through a unified fusion pipeline that aligns various data to support location-based recommendations. Data sources include UK police crime statistics, Ticketmaster event listings, OpenStreetMap points of interest (POIs), online news articles, and routing data. The proposed system comprises four components: a dynamic scoring module for safety and popularity assessment, a conversational AI interface supporting natural language location queries, a personalized recommendation engine informed by user interaction behaviors, and a journey planner that provides users with ranked routes based on safety. The results confirm real-time responsiveness across all four components and contextually meaningful outputs across diverse query types. The key contributions of this work are: (i) a scalable, timely spatial data fusion pipeline for multi-source urban data; and (ii) an AI-augmented decision support framework that promotes more informed and safer urban mobility.

Behavioral Trait Extraction and Clustering from Transit User Text Using Large Language Models

This paper presents a text based frame-work for identifying and analyzing behavioral traits in public transit users using natural language data. Instead of relying on traditional surveys, the approach extracts behavior patterns directly from user generated text, such as reviews or social media posts. Each sample is represented by a hybrid embedding system, which includes meaning for user language and their nuances and behavioral characteristics like urgency, safety sensitivity, and cost awareness is inferred from the system. The representations are clustered and distinct traveler personas are revealed, depending on their approach to time, comfort, safety, and cost. The results show that language based modeling is capable of detecting behavioral differences, nuances and producing relevant traveler profile information without the need for any explicit user data. This demonstrates the potential for behavior analysis to provide a framework for understanding travel behavior at the population level, as well as at the individual level. Furthermore, the results have important implications for the development of behavior based trip planning and personal travel assistance services that respond to the underlying mindsets of users, helping to create more human centric transportation systems.