產品評論於品牌聯想之研究：萃取方法的開發以及評估分析

Abstract

品牌聯想是構成消費者品牌價值模型中很重要的元素，認清品牌聯想對於企業以及消費者雙方都有極大的好處，比如企業可以更瞭解自身品牌在消費者心中的定位，藉以更精準行銷；而消費者也能參考品牌聯想選擇更符合自身需要的商品。過去要清楚了解品牌聯想，需要經過費時又費工的問卷訪談程序，但隨著網路的世代來臨，許多網站成了消費者提供意見的平台，這些累積的使用者生成內容，逐漸成為萃取品牌聯想的資料來源。 過去品牌聯想的相關研究，其目的以及強調的重點非常多樣化。有些著重在比較從不同管道取得的品牌聯想的一致性；有些利用品牌聯想構建網絡，並利用網絡分析相關的變數，找出相對重要的關鍵聯想。而這些研究的共通點是其使用相對簡單的方式萃取品牌聯想，並且決定品牌聯想的權重也相對單純，在評估的階段也沒有使用系統化的方式。 在本篇研究中，我們將三種常用於文字探勘領域的方法應用在品牌聯想的萃取以及權重表示兩大部分，同時也更嚴謹制定篩選的門檻。在評估階段，有別於以往使用人工判斷，我們使用統計和預測模型，在結果中也證實模型的效用會受到品牌聯想相關變數的影響，而利用我們提出的方法所萃取、加權重和篩選過後的品牌聯想對於模型有較好的代表性。

Brand associations are considered as the core component of consumer-based brand equity (CBBE). Identifying brand associations are beneficial to both companies and consumers. Traditionally, the process of eliciting brand associations from consumers is often time and labor consuming. However, as the Internet embeds in people’s lives, consumers start to share everything online, which is called the user generated data (UGC). It gradually becomes one of the data sources from which brand associations can be extracted. The objectives of prior related studies are very diverse. Some focus on the relationship between the extracted associations with the one identified by the company. Some combine network analysis to find the most representative brand associations. Nevertheless, the commonality of these research is that the selection and representation methods performed by which are very preliminary and naive. In our research, we propose using TFIDF, TFICF, and NGD methods to extract and represent brand associations. We conduct both statistical and predictive models as the evaluation approach and prove the model including the variables derived from the brand associations, which are produced by our method attain higher effectiveness.