在現今這個資訊網路、社群網站發展蓬勃的年代，使用者可以在網站上發表自己對每間餐廳的評論。Google Map是一個能夠讓使用者查詢餐廳且具有導航功能，也能夠在該網站留下意見評論反饋的網站。在目前，台灣基於評論資料進行情緒分析的推薦系統較少，於是本研究想要藉由Google Map上的使用者，對於餐廳的評論資料來建立一個推薦系統。本研究指定的餐廳資料範圍為台南市餐廳，以11大種類作為分別，這11種餐廳分別為牛肉湯餐廳、台/中式餐廳、日式餐廳、西式餐廳、美式餐廳、早餐/早午餐店、吃到飽形式餐廳、飲料店、咖啡店、甜點店、酒吧/餐酒館。這些餐廳的種類大致包含一般人對於進食早、午、晚餐及宵夜的需求，本研究會針對每一種類的餐廳製作一個適合該種類餐廳的推薦系統。
推薦系統的核心是找到使用者的偏好，然後針對使用者個人的偏好做出個人化的推薦。在傳統的做法，系統會根據使用者直接的評分來做為喜好的分析依據。然而，現今許多網路平台上，除了原有的評分功能，還會具有讓使用者留下文字說明的功能，這些使用者留下的文字只能夠讓其他的使用者在對一間餐廳行前的參考，這些推薦系統沒有將使用者留下的文字資訊做為評分的依據。本系統之情緒分析部分，即是將這些使用者留下的文字資訊做情緒分析，分析後再將其分析結果用於推薦系統之中。
本研究蒐集了Google Map上台南市11大種類的餐廳評論資料做為本研究的分析資料，共1105960筆，評分資訊以1至5星做為尺度。在情緒分析方面，本研究將每一種類的餐廳取出原始5星評論做為正向資料集，原始3星評論做為中等資料集，原始1星評論做為負向資料集。本研究使用了3種應用於情緒分析之深度學習模型，這三種深度學習模型分別為時間循環神經網路(RNN)的長短期記憶(LSTM)模型、變換器(Transformers)的雙向編碼器表示技術(BERT)模型以及同為BERT家族的小型A Lite BERT(ALBERT)模型，本研究利用這3種模型來分析使用者留下的文字資訊分別分類為正向、中等、負向評論。分析完畢之後結合奇異值分解(Singular Value Decomposition, SVD)、貝氏個人化推薦(Bayesian Personalized Ranking, BPR)以及加權貝氏個人化推薦(Weighted Bayesian Personalized Ranking, WBPR) 3種推薦系統演算法來製成推薦系統。
接著，本研究使用4種推薦系統評估指標來評估此系統的效能，這4種評估指標分別為Precision、MAP(Mean Average Precision)、NDCG(Normalized Discount Cumulative)、Recall。
最後本研究針對洗評論現象做研究，將其中一個資料集的洗評論資料刪除，與不刪除洗評論資料的資料集來做情緒分析及推薦系統，最後再用4種推薦系統評估指標來評估兩者之間的差異。

In today's era of vigorous development of information networks and social networking sites, users can post their own comments on each restaurant on the website. Google Map is a website that allows users to search for restaurants with navigation functions, and also to give feedback on the website. At present, there are few recommender systems in Taiwan for sentiment analysis based on review data, so this study intends to use Google Map restaurant reviews to build a restaurant recommendation system. The scope of restaurant data specified in this study is restaurants in Tainan City, which divided into 11 categories. The types of these restaurants generally include the needs of people for eating breakfast, lunch, dinner and supper. This research will make a recommendation system suitable for each type of restaurant.
This study collected 11 categories of restaurant review data in Tainan City on Google Map as the analysis data for this study, with a total of 1,105,960 reviews. In terms of sentiment analysis, this study takes the original 5-star reviews of each type of restaurant as a positive data set, the original 3-star reviews as a moderate data set, and the original 1-star reviews as a negative data set. LSTM, BERT and ALBERT these three machine learning models were used in this study for sentiment analysis. After the analysis is completed, the recommendation system was built through three algorithms of SVD, BPR and WBPR. Next, The performance of proposed system is evaluated by precision, recall, MAP, NDCG.
Finally, we also discussed the impact of the phenomenon of commercial benefit review comments. The original comment data set and the deleted comment data set are used for sentiment analysis and recommendation system. Finally, four recommendation system evaluation indicators are used to evaluate the difference between the two.

[1] Rushlene Kaur Bakshi, Navneet Kaur, Ravneet Kaur and Gurpreet Kaur, "Opinion mining and sentiment analysis," 2016 3rd International Conference on Computing for Sustainable Global Development (INDIACom), 2016, pp. 452-455.
[2] Sepp Hochreiter and Jürgen Schmidhuber, “Long short-term memory,” Neural computation, vol. 9, no. 8, pp. 1735–1780, 1997.
[3] Tom Young, Devamanyu Hazarika, Soujanya Poria and Erik Cambria, "Recent Trends in Deep Learning Based Natural Language Processing [Review Article]," in IEEE Computational Intelligence Magazine, vol. 13, no. 3, pp. 55-75, Aug. 2018, doi: 10.1109/MCI.2018.2840738.
[4] Sinno Jialin Pan and Qiang Yang, "A Survey on Transfer Learning," in IEEE Transactions on Knowledge and Data Engineering, vol. 22, no. 10, pp. 1345-1359, Oct. 2010, doi: 10.1109/TKDE.2009.191.
[5] Ashish Vaswani et al., “Attention is all you need,” Advances in neural information processing systems, vol. 30, 2017.
[6] Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova, “Bert: Pre-training of deep bidirectional transformers for language understanding,” arXiv preprint arXiv:1810.04805, 2018.
[7] Koustuv Sinha, Robin Jia, Dieuwke Hupkes, Joelle Pineau, Adina Williams, and Douwe Kiela, “Masked language modeling and the distributional hypothesis: Order word matters pre-training for little,” arXiv preprint arXiv:2104.06644, 2021.
[8] Wei Shi and Vera Demberg, “Next sentence prediction helps implicit discourse relation classification within and across domains,” in Proceedings of the 2019 conference on empirical methods in natural language processing and the 9th international joint conference on natural language processing (EMNLP-IJCNLP), 2019, pp. 5790–5796.
[9] Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, and Radu Soricut, “Albert: A lite bert for self-supervised learning of language representations,” arXiv preprint arXiv:1909.11942, 2019.
[10] Charu C. Aggarwal et al. , Recommender systems, vol. 1. Springer, 2016.
[11] Prem Melville and Vikas Sindhwani, “Recommender systems.,” Encyclopedia of machine learning, vol. 1, pp. 829–838, 2010.
[12] Yifan Hu, Yehuda Koren and Chris Volinsky, "Collaborative Filtering for Implicit Feedback Datasets," 2008 Eighth IEEE International Conference on Data Mining, 2008, pp. 263-272, doi: 10.1109/ICDM.2008.22.
[13] Yehuda Koren, Robert Bell, and Chris Volinsky, “Matrix factorization techniques for recommender systems,” Computer, vol. 42, no. 8, pp. 30–37, 2009.
[14] Francesco Ricci, Lior Rokach, and Bracha Shapira, “Introduction to recommender systems handbook,” in Recommender systems handbook, Springer, 2011, pp. 1–35.
[15] Steffen Rendle, Christoph Freudenthaler, Zeno Gantner, and Lars Schmidt-Thieme, “BPR: Bayesian personalized ranking from implicit feedback,” arXiv preprint arXiv:1205.2618, 2012.
[16] Andrew Ian Schein, Alexandrin Popescul, Lyle H. Ungar, and David M. Pennock, “Methods and metrics for cold-start recommendations,” in Proceedings of the 25th annual international ACM SIGIR conference on Research and development in information retrieval, 2002, pp. 253–260.
[17] Lei Guo, Haoran Jiang, Xinhua Wang, and Fangai Liu, “Learning to recommend point-of-interest with the weighted Bayesian personalized ranking method in LBSNs,” Information, vol. 8, no. 1, p. 20, 2017.
[18] 洪梓軒, "餐廳推薦系統結合評論情緒分析", 國立成功大學資訊工程研究所碩士論文, 2018.
[19] 張易筠, "應用BERT語言模型於顧客評論之多面向情緒分析", 國立高雄科技大學資訊管理系碩士論文, 2022.
[20] Xiwang Yang, Yang Guo, Yong Liu, and Harald Steck, “A survey of collaborative filtering based social recommender systems,” Computer communications, vol. 41, pp. 1–10, 2014.
[21] Pengcheng He, Xiaodong Liu, Jianfeng Gao, and Weizhu Chen, “Deberta: Decoding-enhanced bert with disentangled attention,” arXiv preprint arXiv:2006.03654, 2020.
[22] WENGSYX. WENGSYX/Deberta-Chinese-Large. [Online]. Available: https://huggingface.co/WENGSYX/Deberta-Chinese-Large .(accessed on 17 August 2022)
[23] WENGSYX, “CirBERTa: Apply the Circular to the Pretraining Model”. [Online]. Available: https://github.com/WENGSYX/CirBERTa. (accessed on 04 January2023)
[24] Yuanfu Lu, Yuan Fang, and Chuan Shi, ‘Meta-learning on heterogeneous information networks for cold-start recommendation’, in Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2020, pp. 1563–1573.
[25] Timothy Hospedales, Antreas Antoniou, Paul Micaelli and Amos Storkey, "Meta-Learning in Neural Networks: A Survey," in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 9, pp. 5149-5169, 1 Sept. 2022, doi: 10.1109/TPAMI.2021.3079209.
[26] Mingxiao An, Fangzhao Wu, Chuhan Wu, Kun Zhang, Zheng Liu, and Xing Xie, ‘Neural news recommendation with long-and short-term user representations’, in Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 2019, pp. 336–345.
[27] Dawen Liang, Rahul G. Krishnan, Matthew D. Hoffman, and Tony Jebara, ‘Variational autoencoders for collaborative filtering’, in Proceedings of the 2018 world wide web conference, 2018, pp. 689–698.
[28] Diederik P. Kingma and Max Welling, ‘Auto-encoding variational bayes’, arXiv preprint arXiv:1312. 6114, 2013.
[29] Hong-Jian Xue, Xin-Yu Dai, Jianbing Zhang, Shujian Huang, and Jiajun Chen, ‘Deep matrix factorization models for recommender systems’, in IJCAI, 2017, vol. 17, pp. 3203–3209.
[30] Po-Sen Huang, Xiaodong He, Jianfeng Gao, Li Deng, Alex Acero, and Larry Heck, ‘Learning deep structured semantic models for web search using clickthrough data’, in Proceedings of the 22nd ACM international conference on Information & Knowledge Management, 2013, pp. 2333–2338.
[31] Lei Zhang, Shuai Wang, and Bin Liu, ‘Deep learning for sentiment analysis: A survey’, Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, vol. 8, no. 4, p. e1253, 2018.