A Joint Learning Framework for Restaurant Survival Prediction and Explanation

Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 3285 - 3297 December 7-11, 2022 ©2022 Association for Computational Linguistics A Joint Learning Framework for Restaurant Survival Prediction and Explanation Xin Li 1, Xiaojie Zhang 1, Jiahao Peng 1, Rui Mao 1, Mingyang Zhou 1, Xing Xie 2, Hao Liao 1∗ Shenzhen University, China1 Microsoft Research Asia2 {1910273046, 1800271040, 2070276145}@email.szu.edu.cn {mao, zmy, haoliao}@szu.edu.cn xing.xie@microsoft.com Abstract The bloom of the Internet and the recent break- throughs in deep learning techniques open a new door to AI for E-commence, with a trend evolved from using a few financial factors such as liquidity and profitability to using more ad- vanced AI techniques to process complex and multi-modal data. In this paper, we tackle the practical problem of restaurant survival predic- tion. We argue that traditional methods ignore two essential aspects, which are very helpful for the task: 1) modeling customer reviews and 2) jointly considering status prediction and result explanation. Thus, we propose a novel joint learning framework for explainable restaurant survival prediction based on the multi-modal data of user-restaurant interactions and users’ textual reviews. Moreover, we design a graph neural network to capture the high-order inter- actions and design a co-attention mechanism to capture the most informative and meaningful signal from noisy textual reviews. Our results on two datasets show a significant and consis- tent improvement over the SOTA techniques (average 6.8% improvement in prediction and 45.3% improvement in explanation). 1 Introduction Business survival prediction is a hot topic in man- agement and finance literature. Traditional meth- ods rely heavily on financial factors to research(e.g., liquidity, solvency, and profitability)(Ziman, 1991; Lussier, 1996; Pereira et al., 2020). However, there are two significant drawbacks: 1) the financial fac- tors of a shop/company are hard to obtain due to privacy issues; 2) meanwhile, financial factors are macro indicators that reveal the status of a business only on a coarse level. With the development of in- formation techniques, much restaurant-related data can be collected online. For example, people can post check-ins after consuming in a restaurant, and they can share reviews to show how/why they like the restaurant via an online review platform, such as Yelp.com. Moreover, reviews contain informa- tive users’ feedback on a fine-grain level. More importantly, the feedback which deeply reflects the restaurant’s operating status, can in turn help to generate explainable prediction reasons. Some re- cent research works also verify this, and the use of online reviews to understand business perfor- mance is an emerging trend(Babi´c Rosario et al., 2016)(Kong et al., 2017). Recent advances in deep learning have various models that research reviews and interactions for different kinds of tasks, such as recommendation (Wang et al., 2019), fake news detection (Potthast et al., 2018; Wang, 2017), rating prediction (Tay et al., 2018), but little attention has been paid to the application of restaurant survival analysis. In this paper, we propose a novel joint learning framework to tackle the challenging task of explainable restau- rant survival prediction. Our …

Content truncated to 500 words