Cross-Media Recommendation: A Multimodal Approach in Vector Recovery Systems

• Pedro Paulo Carvalho Vieira SENAI Fatesg College
• Rodrigo Santana Esperidião SENAI Fatesg College
• Luiz Felipe Alves Pinheiro SENAI Fatesg College
• Willgnner Ferreira Santos SENAI Fatesg College
• Gustavo Siqueira Vinhal SENAI Fatesg College

DOI: https://doi.org/10.5753/eniac.2026.31249

Resumo

Traditional recommendation systems operate in isolated domains, reinforcing filter bubbles that limit content discovery. This work proposes a content-based cross-media approach connecting films, series, games, anime, and manga in a unified semantic space. Multimodal representations were extracted from textual metadata and cover images using the SigLIP 2 vision-language model. The architecture combines Filtered Vector Search with a score boosting mechanism that re-ranks candidates by weighting popularity scores toward the user’s consumption niche. Empirical evaluation over 20,993 latent vectors showed the multimodal approach outperforms purely textual and visual encodings in cross-media retrieval, achieving an inter-category Recall@10 of 0.7545.

Referências

Barkan, O. and Koenigstein, N. (2017). Item2vec: Neural item embedding for collaborative filtering.

Chronis, Y., Caminal, H., Papakonstantinou, Y., Özcan, F., and Ailamaki, A. (2025). Filtered vector search: State-of-the-art and research opportunities. Proceedings of the VLDB Endowment, 18(12):5488–5492.

Fernández-Tobías, I., Cantador, I., Kaminskas, M., and Ricci, F. (2012). Cross-domain recommender systems: A survey of the state of the art. Proceedings of the 2nd Spanish Conference on Information Retrieval.

Hou, Y., Mu, S., Zhao, W. X., Li, Y., Ding, B., and Wen, J.-R. (2022). Towards universal sequence representation learning for recommender systems.

Malkov, Y. A. and Yashunin, D. A. (2016). Efficient and robust approximate nearest neighbor search using hierarchical navigable small world graphs. CoRR, abs/1603.09320.

Pariser, E. (2011). The Filter Bubble: What the Internet Is Hiding from You. Penguin Press, New York.

Reimers, N. and Gurevych, I. (2019). Sentence-bert: Sentence embeddings using siamese bert-networks. CoRR, abs/1908.10084.

Sassi, I. B. (2025). Scientiarec: a scientific article recommendation system with llm-driven feature extraction. In Proceedings of the 2nd Workshop on Evaluating and Applying Recommendation Systems with Large Language Models (EARL 2025), Prague, Czech Republic. Jointly held with the 19th ACM Conference on Recommender Systems (RecSys 2025).

Sriratanawilai, S., Thongwat, J., Chumpu, R., Payoungkhamdee, P., Nutanong, S., and Limkonchotiwat, P. (2025). Distilling multilingual vision-language models: When smaller models stay multilingual.

Tasenţe, T. (2025). Understanding the dynamics of filter bubbles in social media communication: A literature review. Vivat Academia, pages 1–21.

Tschannen, M., Gritsenko, A., Wang, X., Naeem, M. F., Alabdulmohsin, I., Parthasarathy, N., Evans, T., Beyer, L., Xia, Y., Mustafa, B., Hénaff, O., Harmsen, J., Steiner, A., and Zhai, X. (2025). Siglip 2: Multilingual vision-language encoders with improved semantic understanding, localization, and dense features.

van der Maaten, L. and Hinton, G. (2008). Visualizing high-dimensional data using t-sne. Journal of Machine Learning Research, 9(nov):2579–2605. Pagination: 27.

Vechtomova, O. (2009). Introduction to information retrieval christopher d. manning, prabhakar raghavan, and hinrich schütze (stanford university, yahoo! research, and university of stuttgart) cambridge: Cambridge university press, 2008, xxi+482 pp; hardbound, isbn 978-0-521-86571-5. Computational Linguistics, 35:307–309.

Wu, S., Conneau, A., Li, H., Zettlemoyer, L., and Stoyanov, V. (2020). Emerging cross-lingual structure in pretrained language models.

Zhang, Y., Xie, L., Qiu, R., Liu, J., and Wang, S. (2025). Beyond static llm policies: Imitation-enhanced reinforcement learning for recommendation.