Zhiqing Sun 等人 [11] 指出,使用不恰當的評測方式,可能會造成實驗結果虛高。在進行效果評測時,可能會有一些三元組得分和正確答案的得分相同,這時候針對這些相同得分的結果集合 可以有不同的排序策略:
值得注意的是,TOP 的排序策略在某些情況下是不恰當的,會導致結果虛高。一個直觀的例子是,讓一個模型對所有的候選三元組打分都輸出0,這時正確答案和所有候選三元組得分相同,在 TOP 排序策略下,該模型所有正確答案都排名第一,評測性能達到頂峰。但實際上,由於候選三元組得分相同,該模型在預測時無法判別哪個為正確答案。由此可見,TOP 的排序策略會導致結果虛高。文章指出,RANDOM 的排序策略是最為合理的 [11],這和我們實際的認知也是相符合的,即給定多個相同得分的候選者,我們通常採用隨機的方式來挑選。採用統一的 RANDOM 策略進行評測之後,可以發現有些模型尤其是部分神經網絡模型的性能顯著下降,這表明有些模型由於不恰當的評測方式,導致了結果的虛高,這是我們需要注意甄別的。
[1] Zhanqiu Zhang, Jianyu Cai, Yongdong Zhang, and Jie Wang. 2020. Learning Hierarchy-Aware Knowledge Graph Embeddings for Link Prediction. In AAAI.
[2] Antoine Bordes, Nicolas Usunier, Alberto Garcia-Durán, Jason Weston, and Oksana Yakhnenko. 2013. Translating Embeddings for Modeling Multi-relational Data. In NIPS.
[3] Yankai Lin, Zhiyuan Liu, Maosong Sun, Yang Liu, and Xuan Zhu. 2015. Learning Entity and Relation Embeddings for Knowledge Graph Completion. In AAAI.
[4] Zhiqing Sun, Zhi-Hong Deng, Jian-Yun Nie, and Jian Tang. 2019. RotatE: Knowledge Graph Embedding by Relational Rotation in Complex Space. In ICLR.
[5] Maximilian Nickel, Volker Tresp, and Hans-Peter Kriegel. 2011. A threeway model for collective learning on multi-relational data. In ICML.
[6] Bishan Yang, Scott Wen-tau Yih, Xiaodong He, Jianfeng Gao, and Li Deng. 2015. Embedding Entities and Relations for Learning and Inference in Knowledge Bases. In ICLR.
[7] Maximilian Nickel, Lorenzo Rosasco, and Tomaso Poggio. 2016. Holographic Embeddings of Knowledge Graphs. In AAAI.
[8] Théo Trouillon, Johannes Welbl, Sebastian Riedel, Éric Gaussier, and Guillaume Bouchard. 2016. Complex Embeddings for Simple Link Prediction. In ICML.
[9] Xin Luna Dong, Evgeniy Gabrilovich, Geremy Heitz, Wilko Horn, Ni Lao, Kevin Murphy, Thomas Strohmann, Shaohua Sun, and Wei Zhang. 2014. Knowledge vault: A webscale approach to probabilistic knowledge fusion. In SIGKDD.
[10] Tim Dettmers, Pasquale Minervini, Pontus Stenetorp, and Sebastian Riedel. 2018. Convolutional 2d knowledge graph embeddings. In AAAI.
[11] Daniel Ruffinelli, Samuel Broscheit, and Rainer Gemulla. 2020. You CAN Teach an Old Dog New Tricks! On Training Knowledge Graph Embeddings. In ICLR.
[12] Zhiqing Sun, Shikhar Vashishth, Soumya Sanyal, Partha Pratim Talukdar, and Yiming Yang. 2020. A Re-evaluation of Knowledge Graph Completion Methods. In ACL.
[13] Quan Wang, Zhendong Mao, Bin Wang, and Li Guo. 2017. Knowledge Graph Embedding: A Survey of Approaches and Applications. In TKDE.
作者簡介:陳佳俊,2019年畢業於廈門大學自動化系,獲得工學學士學位。現於中國科學技術大學電子工程與信息科學系的 MIRA Lab 實驗室攻讀研究生,師從王傑教授。研究興趣包括知識表示與知識推理。
五個角度分別為:1)在複雜QA中利用知識圖譜進行神經推理(Neural Reasoning for Complex QA with KGs)2)知識圖譜增強的語言模型(KG-augmented Language Models)3)知識圖譜嵌入:循序推理和歸納推理(KG Embeddings: Temporal and Inductive Inference)