網路口碑之文字數量、文本情感、熵值與結構於產品銷售績效之影響─以電影產業為例

Abstract

過去有不少研究旨在分析網路口碑之於電影票房的影響，常以口碑的數量和評價、及其他電影特徵資料作為票房預測模型之主要自變數。其中口碑數量多取自電影網站上之評論篇數，口碑評價則取電影網站於該電影之統計評分；少有研究探討口碑的文字內容與文本情感之於銷售績效的影響。這份研究旨在探討口碑文字的力量，透過文本挖掘與資料探勘方法，分析口碑的文字數量、文本情感、熵值與結構之於電影銷售績效的解釋能力。 此研究針對口碑文字之影響力以及傳統的口碑數量及評價之影響力進行比較分析，在不考慮其他電影變數的情況下，研究結果指出口碑之文字數量與文本情感不僅能夠取代，甚至超越了傳統的評論數量與統計評價之於電影票房的預測能力，其參數之顯著水準更高，模型之整體解釋能力也更強。 此外，研究發現口碑的文本情感熵值（字群亂度）之於電影票房有負向影響。具體而言，當特定的正面或負面字群比例減少、與其他特定正面或負面字群比例增加之共同作用下，將有助於電影銷售績效之提升。

A lot has been studied in the influences of WOM volume and valence on box office revenues by applying online review quantities and ratings. However, little is known about the force of the quantities and sentiments of words in WOM. This research is conducted to explore the force of words in WOM. With text mining and sentiment analysis techniques, the study is to examine the explanatory power of words to box office performances by word volumes, sentiment scores, and sentiment structures of words, including the entropy and the composition. The study compares the effects of words and of the classical review volume and valence without involving other variables of the movies. The result indicates that the volume and valence of words can substitute, or even surpass, those of reviews by both higher significant level and stronger explanatory power. In addition, the study has found that the entropy of text sentiment (score clusters) has negative effects on box office performances. Specifically, the decrease in the ratios of certain negative or positive word clusters, along with the increase in the ratios of certain negative or positive word clusters, collaboratively generate a positive synergy on box office performances.