<p>This study presents a personalized travel recommender system (TRS) for the Malayalam language, addressing the challenges of recommendation generation in low-resource linguistic contexts. The system utilizes user-generated travelogues and social media reviews to deliver customized travel destination suggestions. A dataset comprising 16,000 unstructured posts from 6,447 travelers was collected and preprocessed to extract key contextual features, including travel mode, travel type, location, and climate preference. The proposed framework employs a multi-model recommendation strategy combining clustering-based, latent factor-based, and rule-based techniques. In the clustering phase, collaborative filtering using K-Means achieved an accuracy of 91%, and content-based filtering using hierarchical agglomerative clustering (HAC) reached 85%. In the latent factor phase, a Hybrid Matrix Factorization (HMF) model attained 82.96% accuracy, outperforming traditional Singular Value Decomposition (SVD) (66.34%). A set of rule-based heuristics-including popularity and feature-driven scoring-was evaluated to enhance explainability, achieving F1-scores up to 0.145. Comparative analysis across all the recommendation engines that were developed demonstrates the superiority of hybrid and clustering-based approaches in balancing accuracy, precision, and recall. The proposed TRS thus establishes an effective framework for personalized, scalable, and explainable recommendations in Malayalam while contributing to sustainable tourism through the promotion of lesser-known destinations in Kerala.</p>