基於多向量融合與 Transformer 用戶偏好建模的個性化新聞推薦

簡易檢索 / 詳目顯示

回結果列表

研究生：	吳昇航 Sheng-Hang Wu
論文名稱：	基於多向量融合與 Transformer 用戶偏好建模的個性化新聞推薦 Personalized News Recommendation via Multi-Vector Fusion and Transformer-based User Preference Modeling
指導教授：	陳彥良 Yen-Liang Chen
口試委員:
學位類別：	碩士 Master
系所名稱：	管理學院 - 資訊管理學系 Department of Information Management
論文出版年：	2025
畢業學年度：	113
語文別：	中文
論文頁數：	57
中文關鍵詞：	新聞推薦、多向量融合、個性化推薦、Transformer 、對比學習、用戶建模
外文關鍵詞：	News Recommendation, Personalized Recommendation, Multi-vector Fusion, Transformer, User Preference Modeling, Contrastive Learning
相關次數：	點閱：10 下載：0
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

在數位資訊爆炸的時代，新聞推薦系統扮演著協助用戶過濾資訊與提升閱讀體驗的關鍵角色。然而，現有方法在新新聞推薦情境中仍面臨冷啟動、語義理解不足及內容同質化等挑戰。傳統內容過濾、協同過濾與知識圖譜技術雖具一定效果，卻難以同時兼顧即時性、多樣性與語義精準度。
為解決上述問題，本文提出一套結合大型語言模型（LLM）與多模態特徵融合的新聞推薦架構，將每篇新聞分解為三種語意向量：由圖片推理視覺吸引力的「吸引力向量」、透過對比學習對齊摘要與 LLM 生成全文的「內容向量」，以及推理多層次分類標籤的「支援向量」。三向量融合後，輸入單層 Transformer 結合用戶歷史行為進行建模與偏好學習，生成個性化推薦結果。
實驗以 MIND-large 資料集為基礎，與 DFM、DKN、NAML、NRMS、LSTUR 及GNewsRec 等主流模型進行比較，結果顯示本模型在 nDCG、Hit@K、MRR 與 AUC 等各項指標上皆具顯著優勢，特別在處理冷啟動新聞上表現突出。進一步的敏感度與消融實驗亦驗證三向量設計具良好穩定性與模組互補性。
綜上，本文的主要貢獻包括：（1）首度結合視覺心理學與 LLM 進行圖片吸引力建模；（2）融合生成式語言能力與對比學習以強化新聞語義表徵；（3）透過語義分類提升推薦準確性與多樣性；（4）整合多模態表示與 Transformer 用戶建模，有效因應冷啟動與同質化挑戰。研究結果展現 LLM 在智慧新聞推薦中整合語義、視覺與語境資訊的潛力，具高度學術與應用價值。

In the era of digital information overload, news recommendation systems play a crucial role in helping users filter vast amounts of information and enhancing their reading experience.However, existing news recommendation methods still face several challenges, particularly in
recommending newly published news items, including cold-start issues, insufficient semantic understanding, and recommendation homogeneity. Traditional approaches based on content filtering, collaborative filtering, or knowledge graphs offer partial solutions, yet often fall short
in simultaneously addressing the timeliness, diversity, and semantic precision required by modern news recommendation.
To overcome these limitations, this study proposes a novel news recommendation framework that combines Large Language Models (LLMs) with multi-modal feature fusion.
The core innovation lies in decomposing news information into three distinct vector modules:(1) the Attractive Vector, derived from visual analysis of news images using LLMs to estimate their visual salience and psychological appeal;(2) the Content Vector, enhanced through contrastive
learning by aligning summaries with full-length news content generated by LLMs to capture deep semantic meaning; and(3) the Supporting Vector, constructed by inferring multi-level
news category labels via LLMs to enrich semantic recognition. These three vectors are fused and subsequently fed into a Transformer model to capture user reading behavior. The resulting user embedding is then combined with candidate news vectors to generate personalized recommendation results.
Experiments are conducted on the MIND-large dataset, where news articles are paired with annotated images. The proposed model is compared against several mainstream baselines,
including DFM, DKN, NAML, NRMS, LSTUR, and GNewsRec. Evaluation metrics include nDCG, Hit@K, MRR, and AUC. Results demonstrate that the proposed model outperforms all
baseline methods across all evaluation criteria, with particularly notable advantages in new item recommendation. Furthermore, sensitivity analyses and ablation studies confirm the complementarity and stability of the proposed tri-vector architecture.
In summary, the contributions of this work include:(1) Introducing LLMs to assess image attractiveness based on visual psychology for use in news recommendation;(2) Enhancing semantic representation of news content by combining generative language modeling with contrastive learning;(3) Improving recommendation precision and diversity through semantic category inference; and(4) Addressing cold-start and content redundancy issues via multimodal fusion and Transformer-based user modeling. This study demonstrates the potential of large language models in unifying semantic, visual, and categorical knowledge within multimodal recommendation systems, offering significant academic and practical value for the development of next-generation intelligent news recommendation systems.

摘要    i
Abstract    ii
Table of Contents    iv
List of Figures    vi
List of Tables    vi
  Introduction    1
1    The Importance of News Recommendation    1
2    Challenges in News Recommendation    2
3    The Role of Large Language Models in News Recommendation    4
4    Proposed Method    5
  Related Work    7
1    The Importance of New Items in News Recommendation    7
2    Common Approaches to New Item Recommendation in News Systems    9
3    Positioning of This Study    12
  Methodology    13
1    Overview    13
2    Problem Definition and Notation    13
3    News Embedding Processing Module    14
3.1    Content Vector    15
3.2    Attractive Vector    17
3.3    Supporting Vector    20
3.4    News Feature Vector    21
4    User Similarity-Based Contrastive Recommendation Module    21
4.1    Parameter Definitions and Model Architecture    21
4.2    Embedding Aggregation via Transformer    22
4.3    Recommendation Procedure    23
5    Model Training and Loss Functions    23
5.1    Contrastive Learning Loss    23
5.2    Recommendation Loss Function    25
5.3    Integrated Loss Function    26
  Experiments    26
1    Dataset and Experimental Setup    26
2    Evaluation Metrics    28
3    Baselines and Comparative Methods    29
4    Sensitivity Analysis    33
5    Ablation Study    37
6    Case Study – Illustration of MVFRec Recommendation Process    40
  Conclusion    42
  Future Work    43
  References    44

                                

[1] Karimi, M., Jannach, D., & Jugovac, M. (2018). News recommender systems–Survey and roads ahead. Information Processing & Management, 54(6), 1203-1227.
[2] Wu, C., Wu, F., Huang, Y., & Xie, X. (2023). Personalized news recommendation: Methods and challenges. ACM Transactions on Information Systems, 41(1), 1-50.
[3] Mitova, E., Blassnig, S., Strikovic, E., Urman, A., Hannak, A., de Vreese, C. H., & Esser, F. (2023). News recommender systems: A programmatic research review. Annals of the International Communication Association, 47(1), 84-113.
[4] Lu, Z., Dou, Z., Lian, J., Xie, X., & Yang, Q. (2015, February). Content-based collaborative filtering for news topic recommendation. In Proceedings of the AAAI conference on artificial intelligence (Vol. 29, No. 1).
[5] Han, K. (2020). Personalized news recommendation and simulation based on improved collaborative filtering algorithm. Complexity, 2020(1), 8834908.
[6] Darvishy, A., Ibrahim, H., Sidi, F., & Mustapha, A. (2020). HYPNER: a hybrid approach for personalized news recommendation. Ieee Access, 8, 46877-46894.
[7] Raza, S., & Ding, C. (2022). News recommender system: a review of recent progress, challenges, and opportunities. Artificial Intelligence Review, 1-52.
[8] Xun, J., Zhang, S., Zhao, Z., Zhu, J., Zhang, Q., Li, J., ... & Wu, F. (2021, October). Why do we click: visual impression-aware news recommendation. In Proceedings of the 29th ACM international conference on multimedia (pp. 3881-3890).
[9] Feng, C., Khan, M., Rahman, A. U., & Ahmad, A. (2020). News recommendation systems-accomplishments, challenges & future directions. IEEE Access, 8, 16702-16725.
[10] Liu, D., Bai, T., Lian, J., Zhao, X., Sun, G., Wen, J. R., & Xie, X. (2019). News Graph: An Enhanced Knowledge Graph for News Recommendation. KaRS@ CIKM, 1.
[11] Wu, F., Qiao, Y., Chen, J. H., Wu, C., Qi, T., Lian, J., ... & Zhou, M. (2020, July). Mind: A large-scale dataset for news recommendation. In Proceedings of the 58th annual meeting of the association for computational linguistics (pp. 3597-3606).
[12] Wang, S., Wang, L., Bu, Y., & Huang, T. (2024). CherryRec: Enhancing News Recommendation Quality via LLM-driven Framework. arXiv preprint arXiv:2406.12243.
[13] Yao, Y., Duan, J., Xu, K., Cai, Y., Sun, Z., & Zhang, Y. (2024). A survey on large language model (llm) security and privacy: The good, the bad, and the ugly. High-Confidence Computing, 100211.
[14] Grigsby, J. L., Jewell, R. D., & Zamudio, C. (2023). A picture’s worth a thousand words: Using depicted movement in picture-based ads to increase narrative transportation. Journal of Advertising, 52(4), 594-612.
[15] An, S., Ma, Z., Lin, Z., Zheng, N., Lou, J. G., & Chen, W. (2024). Make your llm fully utilize the context. Advances in Neural Information Processing Systems, 37, 62160-62188.
[16] Wu, C., Wu, F., Ge, S., Qi, T., Huang, Y., & Xie, X. (2019, November). Neural news recommendation with multi-head self-attention. In Proceedings of the 2019 conference on empirical methods in natural language processing and the 9th international joint conference on natural language processing (EMNLP-IJCNLP) (pp. 6389-6394).
[17] Alabduljabbar, R., Almazrou, H., & Aldawod, A. (2023). Context-Aware News Recommendation System: Incorporating Contextual Information and Collaborative Filtering Techniques. International Journal of Computational Intelligence Systems, 16(1), 137.
[18] Lommatzsch, A., Kille, B., & Albayrak, S. (2017, August). Incorporating context and trends in news recommender systems. In Proceedings of the international conference on web intelligence (pp. 1062-1068).
[19] Wang, H., Zhang, F., Xie, X., & Guo, M. (2018, April). DKN: Deep knowledge-aware network for news recommendation. In Proceedings of the 2018 world wide web conference (pp. 1835-1844).
[20] Zheng, G., Zhang, F., Zheng, Z., Xiang, Y., Yuan, N. J., Xie, X., & Li, Z. (2018, April). DRN: A deep reinforcement learning framework for news recommendation. In Proceedings of the 2018 world wide web conference (pp. 167-176).
[21] Kompan, M., & Bieliková, M. (2010). Content-based news recommendation. In E-Commerce and Web Technologies: 11th International Conference, EC-Web 2010, Bilbao, Spain, September 1-3, 2010. Proceedings 11 (pp. 61-72). Springer Berlin Heidelberg.
[22] Wahana, A., Maylawati, D. S., Wiwaha, B. A., Ramdhani, M. A., & Amin, A. S. (2019, December). News recommendation system using collaborative filtering method. In Journal of Physics: Conference Series (Vol. 1402, No. 7, p. 077010). IOP Publishing.
[23] Wen, H., Fang, L., & Guan, L. (2012). A hybrid approach for personalized recommendation of news on the Web. Expert Systems with Applications, 39(5), 5806-5814.
[24] Gabriel De Souza, P. M., Jannach, D., & Da Cunha, A. M. (2019). Contextual hybrid session-based news recommendation with recurrent neural networks. IEEE Access, 7, 169185-169203.
[25] Lian, J., Zhang, F., Xie, X., & Sun, G. (2018, July). Towards Better Representation Learning for Personalized News Recommendation: a Multi-Channel Deep Fusion Approach. In IJCAI (pp. 3805-3811).
[26] Chuhan Wu, Fangzhao Wu, Mingxiao An, Jianqiang Huang, Yongfeng Huang, and Xing Xie. 2019. Neural news recommendation with attentive multi-view learning. In Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI'19). AAAI Press, 3863–3869.
[27] Wang, H., Wu, F., Liu, Z., & Xie, X. (2020, July). Fine-grained interest matching for neural news recommendation. In Proceedings of the 58th annual meeting of the association for computational linguistics (pp. 836-845).
[28] Adaval, R., Saluja, G., & Jiang, Y. (2019). Seeing and thinking in pictures: A review of visual information processing. Consumer Psychology Review, 2(1), 50-69.
[29] Berger, J., Kim, Y. D., & Meyer, R. (2021). What makes content engaging? How emotional dynamics shape success. Journal of Consumer Research, 48(2), 235-250.
[30] Good, K. D. (2017). Listening to pictures: Converging media histories and the multimedia newspaper. Journalism Studies, 18(6), 691-709.
[31] Kjeldsen, J. E. (2015). The rhetoric of thick representation: How pictures render the importance and strength of an argument salient. Argumentation, 29(2), 197-215.
[32] Brantner, C., Lobinger, K., & Wetzstein, I. (2011). Effects of visual framing on emotional responses and evaluations of news stories about the Gaza conflict 2009. Journalism & Mass Communication Quarterly, 88(3), 523-540.
[33] Kirtley, C. (2018). How images draw the eye: an eye-tracking study of composition. Empirical Studies of the Arts, 36(1), 41-70.
[34] An, M., Wu, F., Wu, C., Zhang, K., Liu, Z., & Xie, X. (2019, July). Neural news recommendation with long-and short-term user representations. In Proceedings of the 57th annual meeting of the association for computational linguistics (pp. 336-345).
[35] Hu, L., Li, C., Shi, C., Yang, C., & Shao, C. (2020). Graph neural news recommendation with long-term and short-term interest modeling. Information Processing & Management, 57(2), 102142.

簡易檢索 / 詳目顯示

相關論文