(161)'A PICTURE SPEAKS A THOUSAND WORDS'. CAN A THOUSAND WORDS DRAW A PICTURE? EXPERIMENTATION USING MATLAB

摘要

Motivated by the well-known saying “a picture speaks a thousand words” we investigate whether wecan use computer graphics and work the other way round claiming also that a “thousand words” candraw - or generate - a picture. In particular, we investigate whether we can exploit single words usedto describe a notion and transform them into building blocks of different color in order to produce afamily of images that capture basic characteristics and properties of this particular notion as conceivedby an audience.Our work falls into the general area of data visualization. Current approaches visualize sets of wordsusing either (i)words in a frequency-based manner or (ii)graphs, e.g., trees, that are generatedaccording to conceptual associations among words provided as input. Instead, in this work, wesuggest an innovative approach according to which sets of words describing a notion are directlymapped to abstract images. Words are transformed to image components. We automatically detectconceptual interconnections among words exploiting their structural properties (like, for example, acommon root)and reflect them to colors. Different colors represent different words of the set. Theword frequency is represented by the size of image components. The final image visually represents anotion and is generated by packing image components using efficient computer science algorithms.To this aim, we first convert words to their hexadecimal equivalents. Then, based on them, wegenerate their equivalent rgb colors. Furthermore, we associate each word used to describe a notionwith a rectangle in the Euclidean space. The size of rectangles visualizes the degree to which aparticular word contributes to the description of a notion and is determined by the frequency ofappearance of words in the description. Efficiently packing all colored rectangles we generate a familyof equal-sized images expressing a notion in terms of an abstract drawing.For our implementation we use Matlab, a simple mathematical programming environment not requiringadvanced programming skills which has been widely used in academia and industry by users comingfrom various backgrounds of engineering, science and economics.The scenario we use as our working case comes from the higher education. Students of theDepartment of Cultural Heritage Management and New Technologies of the University of Patras,Greece, were asked to submit a single word that acts as conceptual reference in the context of eachof four courses, namely Discrete Mathematics, Introduction to Algorithms, Topics in Mobile andWireless Networks and Computational Culture. Three languages were used: Greek, English andSpanish. Then, using the approach previously outlined, we produced images that constitute the visualdescription of each course in each language.Our approach stems from our vision for a culture-inspired data analysis and visualization frameworkwhich fruitfully interlaces cultural elements incorporated in language with methods from computerscience and technology.