網路評價視覺化

Abstract

隨著網際網路的發達，愈來愈多人會在網路上發表自己的看法與評論。這些意見代表著被評論的對象在網路上的民意反應，了解這些民意可以讓被評論的對象更瞭解自身的優缺點並加以改進，不論是企業或是知名人物對於這些網路民意的瞭解程度將會是提升自身競爭力的一個關鍵因素。 然而，這些網路意見所反應出來的民意並沒有一個可以具體說明程度的指標，若要一筆一筆瞭解這些網路意見的內容會是一件困難且耗時的任務。在我們的研究中，提出一個反應網路民意的指標：”名聲 (Reputation)”。我們利用自動化意見分析的方法判斷出網路評論內容的正負傾向與意見強度，我們稱之為”情緒分數 (Sentiment score)”。除此之外，我們還提出了一個”認同度 (Support)”的概念，來衡量每筆評論被認同的程度。綜合這兩個數值來衡量一筆意見，再透過加權平均的方法整合這些意見計算一個對象的”名聲”來代表網路上大家對它的看法。接著，我們衡量某個對象的整體網路民意對於他人的影響力。這個影響力的量化是要讓企業或是知名人物對於自身的名聲給予網路上瀏覽者的影響力有一個具體的概念。 為了能讓這些指標更容易被理解，我們提出了資訊視覺化的方法來呈現這些指標。我們利用”泡泡圖 (bubble chart)”可以呈現多維度資訊的特性來呈現這些指標在不同對象之間的差異，並且搭配階層的概念來呈現這些指標在不同階層中的變化。

With the development of Internet, more and more people are sharing their ideas and views via the World Wide Web. These opinions represent the reaction from the public discussions toward the target; the targets can realize their strengths and improve weaknesses through the public opinion. In any circumstances, figuring out the public’s opinions is a key factor to enhancing competitiveness for both corporate groups and celebrities. However, there are no quantified indexes to interpret the opinions, and it is also a complex task to analyze them. Therefore, we proposed an index, called “Reputation” to represent the overall public opinions. The index is an aggregation of the whole public opinion. It consists of two dimensions: “Sentiment score” and “Support”. We used a weighted average method to aggregate opinions, which formed the final reputation of an entity. This entity is the attitudes from the public. In addition, we further measured the impact of the opinions to other people. The goal of quantifying the impact value is to let corporate groups and celebrities realize their own reputation as well as the impact of public opinions. In order to clearly understand this, we proposed a visualization system to display these indexes. We will use the characteristic of Bubble Chart which can simultaneously display multiple dimensions to differentiate multiple targets’ indexes. Moreover, we also apply circle packing display to visualize hierarchy distribution within each entity.