我的研究提出了SpecRank模型，一種用於群眾外包的環境下各種擁有不同專業領域的受測者身上所收集到的物品比較資料找出所有物品排序的演算法。我們假設物品在實際情況下可能分屬於請求者未知的類別，且受測者提供標籤的能力取決於他們對這些類別的熟悉程度。針對以上的問題，我們提出了考量受測者專業程度的機率模型並將之命名為specialty-Bradley Terry Luce (specialty-BTL) 將受測者對於各領域具有不同的專業程度納入考量。此外，我們更將物品的分數表示為物品的特徵向量及其所屬類別特徵向量之間的內積，來描述物品與不同類別之間的關係。而受測者提供的標籤的能力更需取決於他們的專業特徵向量和類別特徵向量之間的相關性。基於實際情況下物品的類別皆為未知，我們採用高斯混合模型，並透過Expectation-Maximization (EM)算法對物品進行排名和辨識類別。另外我們也考慮，當我們擁有一些比較數據並運行SpecRank方法時，此時我們如何利用估計的參數為即將到來的工作者分配合適的任務。

關鍵字 – 排序融合、群眾外包、Variational EM演算法

This work proposes SpecRank, an algorithm for the ranking of objects from pairwise comparisons provided by workers with different specialties in a crowdsourced setting. The objects
may belong to different classes that are unknown to the requester, and the workers' ability to provide reliable labels may depend on their specialty or familiarity with these classes. The specialty-Bradley Terry Luce model (specialty-BTL) is first proposed to incorporate workers' specialty in the quality of their pairwise comparison labels. Then, the scores of objects are modeled as inner products between the latent feature vectors of the objects and their associated classes. The workers' ability to provide accurate pairwise labels is also determined by the correlation between their specialty feature vectors and the class feature
vectors. By adopting a mixture model on the objects' class labels, SpecRank is proposed based on the variational expectation-maximization (EM) algorithm to perform joint ranking and clustering of the objects. In addition, we consider the problem that when we have some comparison data and run our SpecRank method, how we utilize the estimated parameters to assign suitable tasks for the coming worker at this moment.
Keywords - Rank aggregation, Crowdsourcing, Variational EM algorithm

[1] A. Kovashka, O. Russakovsky, L. Fei-Fei, K. Grauman,et al., “Crowdsourcing in computer vision, ”Foundations and TrendsR© in computer graphics and Vision, vol. 10, no. 3,pp. 177–243, 2016.
[2] A. Wang, C. D. V. Hoang, and M.-Y. Kan, “Perspectives on crowdsourcing annotations for natural language processing, ”Language resources and evaluation, vol. 47, no. 1,pp. 9–31, 2013.
[3] M. Lease and E. Yilmaz, “Crowdsourcing for information retrieval,” in ACM SIGIR Forum, vol. 45, pp. 66–75, 2012.
[4] “Amazon Mechanical Turk,” 2001. https://www.mturk.com/.
[5] “Figure Eight,” 2007. https://www.figure-eight.com.
[6] “Clickworker,” 2009. https://www.clickworker.com.
[7] C. Dwork, R. Kumar, M. Naor, and D. Sivakumar, “Rank aggregation methods for the web,” in Proceedings of the 10th international conference on World Wide Web, pp. 613–622, ACM, 2001.
[8] S. Negahban, S. Oh, and D. Shah, “Rank Centrality: Ranking from pairwise comparisons, ”Operations Research, vol. 65, pp. 266–287, 2016.35
[9] X. Chen, P. N. Bennett, K. Collins-Thompson, and E. Horvitz, “Pairwise ranking ag-gregation in a crowdsourced setting,” in Proceedings of the sixth ACM international conference on Web search and data mining, pp. 193–202, 2013.
[10] Y. Liu, W. Li, Y. Lin, and L. Jing, “Spectral geometry for simultaneously clustering and ranking query search results,” in Proceedings of the 31st annual international ACMSIGIR conference on Research and development in information retrieval, pp. 539–546,2008.
[11] J. Li, Y. Baba, and H. Kashima, “Simultaneous clustering and ranking from pairwise comparisons,” in Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence (IJCAI), pp. 1554–1560, 2018.
[12] R. Wu, J. Xu, R. Srikant, L. Massoulie, M. Lelarge, and B. Hajek, “Clustering and inference from pairwise comparisons, ” in Proceedings of the 2015 ACM SIGMETRICS International Conference on Measurement and Modeling of Computer Systems, pp. 449–450, 2015.
[13] A. Salehi Abari and C. Boutilier, “Preference-oriented social networks: Group recommendation and inference, ” in Proceedings of the 9th ACM Conference on Recommender Systems, pp. 35–42, 2015.
[14] I. Cantador, P. Brusilovsky, and T. Kuflik, “2nd workshop on information heterogeneity and fusion in recommender systems (hetrec 2011),” in Proceedings of the 5th ACM conference on Recommender systems, RecSys 2011, 2011.[15] F. Fogel, A. d’Aspremont, and M. Vojnovic, “Serialrank: Spectral ranking using seriation, ” in Advances in Neural Information Processing Systems, 2014.36
[16] L. Baltrunas, T. Makcinskas, and F. Ricci, “Group recommendations with rank aggregation and collaborative filtering,” in Proceedings of the Fourth ACM Conference on Recommender Systems, pp. 119–126, 2010.
[17] Y. Gao and K. Xu, “pRankAggreg: A fast clustering based partial rank aggregation ,”Information Sciences, vol. 478, pp. 408–421, 2019.
[18] L. Cao, X. Jin, Z. Yin, A. Del Pozo, J. Luo, J. Han, and T. S. Huang, “Rankcompete: Simultaneous ranking and clustering of information networks, ”Neurocomputing, vol. 95,pp. 98–104, 2012.
[19] Y. Sun, J. Han, P. Zhao, Z. Yin, H. Cheng, and T. Wu, “RankClus: Integrating clustering with ranking for heterogeneous information network analysis, ” in Proceedings of the 12th International Conference on Extending Database Technology: Advances in Database Technology (EDBT), pp. 565–576, 2009.
[20] R. A. Bradley and M. E. Terry, “Rank analysis of incomplete block designs: I. the method of paired comparisons, ”Biometrika, vol. 39, no. 3/4, pp. 324–345, 1952.
[21] S. Theodoridis, Machine Learning: a Bayesian and Optimization Perspective. AcademicPress, 2015.
[22] U. Von Luxburg, “A tutorial on spectral clustering, ”Statistics and computing, vol. 17,no. 4, pp. 395–416, 2007.
[23] P. Diaconis, “Group representations in probability and statistics, ”Lecture notes-monograph series, vol. 11, pp. i–192, 1988.
[24] M. G. Kendall, “Rank correlation methods.,” 1948.37
[25] E. V. B. P. P. E. Abbasnejad, S. Sanner, “Learning community-based preferences via dirichlet process mixtures of gaussian processes, ” in Proceedings of the 23rd International Joint Conference on Artificial Intelligence (IJCAI), 2013