It is suggested that embedding-based retrieval is more effective in capturing semantic information, including synonym usage and implicit contextual relationships, within the evaluated dataset, and compact embedding models such as MiniLM may provide an alternative approach to traditional keyword-based retrieval methods for small-scale recommendation systems.
Abstract
The massive volume of textual descriptions on hotel booking platforms makes it difficult for recommendation systems to accurately match user preferences. Traditional keyword-based retrieval methods, such as TF-IDF, often struggle to capture semantic relationships when relevant terms do not explicitly overlap. This study evaluates the performance of keyword-based (TF-IDF) and embedding-based (paraphrase-multilingual-MiniLM-L12-v2) retrieval approaches in a content-based hotel recommendation system using a small-scale dataset. The dataset consists of 30 unique Traveloka hotels in Yogyakarta collected from Kaggle, representing a resource-constrained experimental setting. Evaluation was conducted using a black-box approach with 10 dynamic synthetic queries and assessed through Precision@3 (P@3) and Mean Average Precision (MAP). The results indicate that MiniLM achieved higher retrieval effectiveness than TF-IDF, with a mean P@3 of 0.3667 and a mean MAP of 0.1378, compared with 0.3000 and 0.1333, respectively. These findings suggest that embedding-based retrieval is more effective in capturing semantic information, including synonym usage and implicit contextual relationships, within the evaluated dataset. Therefore, compact embedding models such as MiniLM may provide an alternative approach to traditional keyword-based retrieval methods for small-scale recommendation systems.
Recommendation engines are significant tools that provide personalized recommendations especially in case of cold start problem where historical interaction data is insufficient. Deep learning techniques such as the state-of-the-art RAG (Retrieval-Augmented Generation) method make the recommendation procedure better by utilizing the concept of contextual preference. They, however, depend on large language models and therefore it increases the computational cost, slow the process and make it less explainable. This paper discusses a lightweight and explainable RAG-based recommendation engine. In the proposed model, SBERT embeddings and FAISS indexing are used for effective semantic retrieval and hybrid ranking. Explainability component is included in the architecture to provide explanations using similarity score, rating, popularity, author similarity and publication era. Experiments are performed on Kaggle Book-Crossing dataset using ranking metrics such as Recall, NDCG and MRR at various K values, where K is the number of recommendations selected from the ranked list. As per the experimental results, the proposed framework achieves its highest MRR score of 0.1509 at K = 25.
Kanthi Purnima, S. Surekha· International Journal For Mu...· 0 citations
This study validates the effectiveness of the BERT model in semantic similarity calculation, providing more accurate technical support for related application scenarios, and laying the foundation for subsequent model optimization and lightweighting research.
Jiachen Gao· International Conference on...· 0 citations
The increasing number of students and research topics has led to challenges in identifying relevant and high-quality academic references efficiently. Traditional keyword-based search methods often fail to capture the semantic meaning of research topics, resulting in less accurate recommendations. This study proposes a semantic-based recommendation system using Multilingual BERT (mBERT) to compute similarity between user queries and academic publication titles. The system leverages text embeddings and cosine similarity to identify relevant references from the Semantic Scholar database. Experimental results show that the proposed approach achieves an accuracy of 90%, precision of 93.75%, recall of 93.75%, and F1-score of 0.9375 in semantic similarity classification. The results demonstrate that BERT-based semantic analysis effectively improves the relevance of academic reference recommendations compared to traditional approaches.
Paramitha Nerisafitra, Raasikh Fil’Ilmi, A. Qoiriah et al.· E3S Web of Conferences· 0 citations
This work aims to explore how well vector-based pseudo relevance feedback can adapt to dense retrieval models when it is not feasible to use surface-form information to pick discriminating expansion tokens.
X. Velez, Andrew Yates, Eugene Yang et al.· Annual International ACM SIG...· 0 citations
Findings confirm that semantic embedding methods provide superior ranking quality and cross-lingual robustness, offering a scalable and translation-free solution applicable to multilingual product recommendation systems in commercial settings.
Kevin De Rafael Rio Aryanto, Faulinda Ely Nastiti, Ridwan Dwi Irawan· IC-ITECHS· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
