A Deep Reinforced Sequence-to-Set Model for Multi-Label Classification

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 5252–5258 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 5252 A Deep Reinforced Sequence-to-Set Model for Multi-Label Classification Pengcheng Yang1,2, Fuli Luo2, Shuming Ma2, Junyang Lin2, Xu Sun1,2 1Deep Learning Lab, Beijing Institute of Big Data Research, Peking University 2MOE Key Lab of Computational Linguistics, School of EECS, Peking University {yang pc, luofuli, shumingma, linjunyang, xusun}@pku.edu.cn Abstract Multi-label classification (MLC) aims to pre- dict a set of labels for a given instance. Based on a pre-defined label order, the sequence-to- sequence (Seq2Seq) model trained via maxi- mum likelihood estimation method has been successfully applied to the MLC task and shows powerful ability to capture high-order correlations between labels. However, the output labels are essentially an unordered set rather than an ordered sequence. This in- consistency tends to result in some intractable problems, e.g., sensitivity to the label order. To remedy this, we propose a simple but effective sequence-to-set model. The proposed model is trained via reinforcement learning, where re- ward feedback is designed to be independent of the label order. In this way, we can reduce the dependence of the model on the label or- der, as well as capture high-order correlations between labels. Extensive experiments show that our approach can substantially outperform competitive baselines, as well as effectively re- duce the sensitivity to the label order. 1 1 Introduction Multi-label classification (MLC) aims to assign multiple labels to each sample. It can be applied in many real-world scenarios, such as text catego- rization (Schapire and Singer, 2000) and informa- tion retrieval (Gopal and Yang, 2010). Due to the complex dependency between labels, a key chal- lenge for the MLC task is how to effectively cap- ture high-order correlations between labels (Zhang and Zhou, 2014). When involving in capturing high-order correla- tions between labels, one line of research focuses on exploring the hierarchical structure of the la- bel space (Prabhu and Varma, 2014; Jernite et al., 2017; Peng et al., 2018; Singh et al., 2018), while 1The code is available at https://github.com/ lancopku/Seq2Set another line strives to extend specific learning al- gorithms (Zhang and Zhou, 2006; Baker and Ko- rhonen, 2017; Liu et al., 2017). However, most of these work tends to result in intractable computa- tional costs (Chen et al., 2017). Recently, based on a pre-defined label order, Nam et al. (2017); Yang et al. (2018) succeeded in applying the sequence-to-sequence (Seq2Seq) model to the MLC task, which shows its power- ful ability to capture high-order label correlations and achieves excellent performance. However, the Seq2Seq model suffers from some thorny flaws on the MLC task. The output labels are essentially an unordered set with swapping-invariance2, rather than an ordered sequence. This inconsistency usu- ally leads to some intractable problems, e.g., sen- sitivity to the label order. Previous work (Vinyals et al., 2016) has shown that the order has a great impact on the performance of the Seq2Seq model. Therefore, the performance …

Content truncated to 500 words