運用資料探勘於自動化色彩語意分析之研究

色彩是一種設計語言，影響視覺傳達與人們的心理認知（色彩意象）。色彩語意為溝通色彩意象的詞彙，因文化等因素有所差異，過去以心理實驗調查的方式，無法即時回應設計需求的增長。因此本研究提出一個色彩語意分析方法，結合網路大數據、資料探勘與色彩意象尺度，嘗試得出合理之詞彙與色彩對應關係。本研究結果顯示，與過去文獻結果相比較，使用卷積神經網路可得詞彙對應色相，唯有無彩色的部分效果不彰。自然語言辨識結合色彩意象尺度可得詞彙對應色調，部分結果與文獻呈現極端值，顯示若要將色彩意象尺度作為設計標準，仍需經過專家在地化調整。此方法未來可進行參數化與設計應用結合，開發自動化色彩分析系統，並作為其他設計應用研究之參考。

關鍵字

資料探勘；色彩語意；色彩意象

並列摘要

Color plays a crucial role in visual communication and cognition (color image). Color semanticsarewordshelp us to communicate color image,influenced by factors as culture, experience and so on, but psychological experimental investigations could not respond to the growth of design requirementsimmediately. Therefore, we propose a method of color semantic analysis, combining big data, data mining and color image scale, and tries to obtain a reasonable correspondence between words and colors. Compared with the results of literature, the results of this study shows that convolutional neural networks can obtain words corresponding to the hue, only in the achromatic color can not achieve a good identification. Natural language processing combined with the color image scale can obtain words corresponding to the tone, but some results show extreme difference with literature. It shows that if color image scale is used as design standard, it still needslocalization by experts. In the future, this method can be improved by parameterization, to develop an automated color analysis system, as a reference for color design.