喜怒（不）形於色？平台經濟的情感分析

To Be or Not to Be Demonstrative? Toward a Sentiment Analysis of the Platform Economy

本文探討平台經濟做為一種創新的商業模式，如何在社會系統中成長與擴散，我們採用自然語言處理的文本情感分析做為研究方法，量化平台生態系統中消費者的情感變化，以此來理解其過程。我們透過爬蟲程式從Google Play抓取Foodpanda與Uber Eats總共35,909篇用戶評論，時間跨度鎖定在2016年10月至2019年11月，根據樣本大小區分成訓練集與測試集，最後得出每月介於-1和1之間的情感分數。我們指出Foodpanda與Uber Eats剛進入台灣時，社會對於平台經濟的陌生感會轉變為負面情感。接者，因為技術與服務逐漸完善，以及推出許多行銷活動，社會大眾開始接受平台經濟，促使雙方的情感分數逐漸上升，其中又以Foodpanda成長最快。最後，隨著外送平台的市場逐漸成熟，兩者情感分數的上升幅度趨緩。我們進一步提出一個平台經濟的動態發展模式，並提供一些實務上的建議。

This paper aims to explore how platform economy as a new business model diffuses in the modern social system. Methodologically, we use sentiment analysis of natural language processing to measure the changes in consumers' sentiment to understand the dynamic process. We collect Foodpanda and Uber Eats' online user reviews data from Google Play (35,909 in total) by web crawler, ranging from October 2016 to November 2019. Data are further divided into training set and testing set based on the sample size. Then, we compute monthly sentiment scores between -1 and 1. We found that users' unfamiliarity became negative sentiment when Foodpanda and Uber Eats first entered Taiwan. Due to Foodpanda and Uber Eats' increasingly improved service and active promotions, more users began to attend to the platform economy, leading to the increase of their sentiment score - Foodpanda in particular. As the platform market became mature, the increase in their scores slowed down. Furthermore, we develop a four-phase development model of platform economy and offer some practical implications.