在高度競爭且媒體環境快速變遷的市場中，品牌如何有效與消費者建立連結已成為關鍵課題。品牌代言人作為重要的行銷策略之一，能透過其知名度與形象影響消費者認知，提升品牌辨識度與信任感。然而，若代言人形象與品牌定位不一致，可能導致品牌認知混淆甚至負面影響。因此，如何系統性地評估並推薦適合的代言人，成為重要且具挑戰性的研究議題。
傳統代言人選擇多依賴行銷專家的經驗與市場直覺，並輔以調查資料。然而，在社群媒體與數位平台興起後，影響者數量快速增加，使候選集合與資訊結構日益複雜。單一指標已難以全面衡量品牌與代言人之間的契合程度，且此類匹配涉及價值觀、風格與文化意涵等語意層面，使問題具高度複雜性。此外，實務資料常呈現稀疏與不完全觀測的特性，未合作並不代表不適合，進一步增加建模困難。
本研究提出一套結合大型語言模型與對比學習的品牌代言人推薦方法。首先，蒐集品牌與名人的多來源文本資料，透過大型語言模型進行語意摘要，生成品牌形象與名人人設描述，再利用嵌入模型將其轉換為帶有語義之數值向量。接著，透過對比學習建構共享語意空間，使實際合作的品牌–名人配對在向量空間中距離更接近，而不相符組合則被拉遠，以學習其語意契合關係。在推薦階段，系統將品牌轉換為向量表示，並與所有候選名人的向量計算相似度進行排序，產生推薦名單。
實驗結果顯示，本研究方法能有效捕捉品牌與名人之間的語意關聯，在推薦表現上優於基準方法。本研究將品牌代言媒合問題轉化為語意表徵學習與排序推薦任務，提出一種可量化契合度的資料驅動方法，提升代言人選擇的科學性，並拓展行銷科技與推薦系統的應用潛力。未來可進一步整合多模態資料，以提升模型表現與實務價值。

In today’s rapidly evolving media landscape, selecting suitable celebrity endorsers has become a critical yet challenging task for brands. While endorsements can enhance brand recognition and trust, mismatches between brand positioning and celebrity image may lead to negative consumer perceptions. Traditional approaches rely on expert judgment and simple metrics, which are insufficient to capture the complex and semantic nature of brand–celebrity compatibility.
This study proposes a data-driven framework that integrates Large Language Models (LLMs) with contrastive learning for brand endorser recommendation. Multi-source textual data are leveraged to generate semantic descriptions of brand identity and celebrity personas using LLMs, which are then transformed into embedding representations. A contrastive learning approach is employed to construct a shared embedding space, where matched brand–celebrity pairs are drawn closer while mismatched pairs are pushed apart, enabling the model to learn semantic compatibility. During inference, candidate endorsers are ranked based on similarity between brand and celebrity embeddings.
Experimental results demonstrate that the proposed method effectively captures semantic relationships and outperforms baseline approaches. This study formulates endorsement matching as a representation learning and ranking problem, providing a quantifiable and scalable solution for data-driven marketing decision-making.

[1] M. Kosinski, D. Stillwell, and T. Graepel, Private Traits and Attributes Are Predictable from Digital Records of Human Behavior, Proceedings of the National Academy of Sciences, Vol. 110, No.15, pp. 5802-5805, 2013.
[2] Y. Wu, M. Kosinski, and D. Stillwell, Computer-based Personality Judgments Are More Accurate Than Those Made by Humans, Proceedings of the National Academy of Sciences, Vol. 112, No.4, pp. 1036-1040, 2015.
[3] Y. Cao and M. Kosinski, Large Language Models Know How the Personality of Public Figures is Perceived by the General Public, Scientific Reports, Vol. 14, No.1, 6735, 2024.
[4] H. Peters and S. Matz, Large Language Models Can Infer Psychological Dispositions of Social Media Users, Proceedings of the National Academy of SciencesNexus, Vol. 3, No. 6, 2024.
[5] H. Lv, G. Yu, and X. Tian, A Matching Recommendation Algorithm for Celebrity Endorsement on Social Network, 20th Annual International Conference on Management Science and Engineering, 2013.
[6] N. Cai, S. Li, Z. Yu, and M. Shi, Select the Potential Brand Spokesperson on Social Network Based on the Smart Link Prediction Algorithm, Journal of Intelligent & Fuzzy Systems, Vol. 37, No. 1, 2019.
[7] C. Saridakis, C.S. Katsikeas, S. Angelidou, M. Oikonomidou, and P. Pratikakis, Mining Twitter Lists to Extract Brand-Related Associative Information for Celebrity Endorsement, European Journal of Operational Research, Vol. 311, No. 1, 2023.
[8] T. Mikolov, K. Chen, G. Corrado, and J. Dean, Efficient Estimation of Word Representations in Vector Space, International Conference on Learning Representations, 2013.
[9] N. Reimers and I. Gurevych, Sentence-BERT: Sentence Embeddings Using Siamese BERT-networks, 2019 Conference on Empirical Methods in Natural Language Processing, 2019.
[10] F Schroff, D Kalenichenko, and J Philbin, FaceNet: A Unified Embedding for Face Recognition and Clustering, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015.
[11] A. Oord, Y. Li, and O. Vinyals, Representation Learning with Contrastive Predictive Coding, arXiv preprint arXiv:1807.03748, 2018
[12] P. Khosla, P. Teterwak, and C. Wang, Supervised Contrastive Learning, Advances in Neural Information Processing Systems (NeurIPS), 2020.
[13] P. Covington, J. Adams, and E. Sargin, Deep Neural Networks for YouTube Recommendations, 10th ACM Conference on Recommender Systems, 2016.