searching for the rainbow

10
1 Colourful Language: Searching for the Rainbow (poster) 1 Eleanor Maclure 1 MA Graphic Design, University of the Arts London/London College of Communication, [email protected] 1. Introduction The philosopher Wittgenstein famously asked ‘How do I know that this color is red? —It would be an answer to say: I have learnt English.’ [1] But how do we know what is red? In our current, digitised age perhaps a more fitting response would be: ‘Google it’. For many the Internet has become our primary source of information, it is the foundation of our knowledge economy. Largely inclusive and participatory, as an entity it represents the culmination of everything that has been written, posted and uploaded, with each new contribution slightly altering the shape of the whole. But how does Google know what is red? Tags, Search Engine Optimisation (SEO) and complex algorithms now create the precedence for the information we receive when we search for something. Search engines are able to dictate the type of information we consume, influencing our perceptions of what is important with the ranking of their search results. 2. Background This project forms part of a larger investigation into the way we use language to describe colours for my MA Major Project. The use of Google Image Search as a tool to provide a visual interpretation of colour terms began as no more than an exercise in understanding, to enable me to progress with other aspects of my research. I had begun using Google Image Search to provide a visual reference for colour names that I was unfamiliar with, which had been collected at the outset of the investigation. It was exclusively a means to an end and not written into the original proposal for the project. While it proved to be a very useful process, through repeated attempts it became apparent that the search results themselves generated a fascinating body of images, providing an insight into the fluid nature of the Internet, the relationship between the colours in the images and the search terms and the level of concurrence across the colour of the images retrieved by each search. While there are a number of systems available for accurate colour specification, we all still use colour terms rather than precise notation to describe what we see in everyday life. But how well do we use these terms? How consistent and precise are we when it comes to defining what is claret, wine, maroon or burgundy? Would we feel confident asserting that something was beige rather than taupe? Do we know the difference between lilac and lavender? Is there even a difference? Through initial trials for the process it became clear that although they are translated through Google’s algorithms, collectively, the images retrieved from internet searches represent the particular level of understanding of colour and colour terms by those posting them, tagging them or writing the accompanying text.

Upload: eleanor-maclure

Post on 20-Nov-2015

212 views

Category:

Documents


0 download

DESCRIPTION

Paper presented at the IX Conferenza del Colore in Florence, Italy in 2013. Joint meeting of the Colour Group GB and the Italian Gruppo del Colore.

TRANSCRIPT

  • 1

    Colourful Language: Searching for the Rainbow (poster) 1Eleanor Maclure

    1MA Graphic Design, University of the Arts London/London College of Communication, [email protected]

    1. Introduction The philosopher Wittgenstein famously asked How do I know that this color is red? It would be an answer to say: I have learnt English. [1] But how do we know what is red? In our current, digitised age perhaps a more fitting response would be: Google it. For many the Internet has become our primary source of information, it is the foundation of our knowledge economy. Largely inclusive and participatory, as an entity it represents the culmination of everything that has been written, posted and uploaded, with each new contribution slightly altering the shape of the whole. But how does Google know what is red? Tags, Search Engine Optimisation (SEO) and complex algorithms now create the precedence for the information we receive when we search for something. Search engines are able to dictate the type of information we consume, influencing our perceptions of what is important with the ranking of their search results.

    2. Background This project forms part of a larger investigation into the way we use language to describe colours for my MA Major Project. The use of Google Image Search as a tool to provide a visual interpretation of colour terms began as no more than an exercise in understanding, to enable me to progress with other aspects of my research. I had begun using Google Image Search to provide a visual reference for colour names that I was unfamiliar with, which had been collected at the outset of the investigation. It was exclusively a means to an end and not written into the original proposal for the project. While it proved to be a very useful process, through repeated attempts it became apparent that the search results themselves generated a fascinating body of images, providing an insight into the fluid nature of the Internet, the relationship between the colours in the images and the search terms and the level of concurrence across the colour of the images retrieved by each search. While there are a number of systems available for accurate colour specification, we all still use colour terms rather than precise notation to describe what we see in everyday life. But how well do we use these terms? How consistent and precise are we when it comes to defining what is claret, wine, maroon or burgundy? Would we feel confident asserting that something was beige rather than taupe? Do we know the difference between lilac and lavender? Is there even a difference? Through initial trials for the process it became clear that although they are translated through Googles algorithms, collectively, the images retrieved from internet searches represent the particular level of understanding of colour and colour terms by those posting them, tagging them or writing the accompanying text.

  • 2

    3. Methodology There were several stages to the methodology as the project developed and was extended. When it became apparent that Google Image Search could be an effective tool for a visual investigation into colour and language, I felt it was important to devise a structured and rigorous enough methodology while remaining appropriate to the scope and ethos of my design-based course. In this instance a generative approach was highly applicable as Google Image Search could be used as an input/output system. English was the language chosen for the project as I am a native speaker, was using a UK based IP address and internet service provider with Google UK as a default and because English is still the dominant language on the Internet both in terms of content generate and the native language of users. [2] While cross-cultural and linguistic differences in the use of colour terms is a rich area of study I had already stipulated when writing the proposal that the wider investigation would be confined to English colour terms, in order to contain the scope of the project at my particular level of study. To create my initial criteria for the colour terms to use in the search I employed the eleven basic colour terms in English as defined by Berlin & Kay [3]: black, white, red, green, yellow, blue, brown, purple, pink, orange and grey. They offered a clearly defined set of widely used colour terms, representing a spread of both spectral, non-spectral and achromatic colours, that were expected to be well understood by native speakers, and are collectively supported by decades of academic research. Although there have been criticisms of bias in the methodology used by Berlin & Kay and questions over the universality of their theory of the evolution of colour terms, in the absence of any other well established body of basic colour terms I felt that these arguments would not affect the validity of the methodology enough to discount the eleven basic colour terms as a basis for the search. Before beginning the process I completely cleared my browser history and deleted any cookies, so that the results would not be influenced by my internet history. This was followed by inputting each of the eleven basic colour terms into Google Image Search with the addition of the word colour. The inclusion of the word colour in the search terms was an undesirable but an unfortunate consequence of a number of high profile celebrities who feature colour terms as part of their names, such as the rapper Chris Brown and the singer Pink. This created significant distortions in the image results for a small number of the colour terms in the initial trials of the process, without adding anything of value to the comparison. While it would have been preferable to use the colour term alone, the addition of the word 'colour' was deemed a necessary concession. It should also be noted that for consistency I used the English spellings of the words 'grey' and 'colour' throughout, rather than the American English spellings 'gray' and 'color'. Once each search had been completed I then took, without exception, the first thirty images from the search results. When developing the methodology for the project thirty was considered to be a suitable number for the study, as it would generate a reasonably sized body of images without creating an excessive time burden for this aspect of the investigation. These images would both represent the original result of

  • 3

    each colour term search and provide the basis for analysis using image manipulation processes. During the development of my proposal for the wider project I had trialled a number different digital manipulation techniques as a way of visually analysing the colour content of each image. These included several processes applied in Photoshop and using Colourphon, an online colour indexing system [4]. For the body of images resulting from Google Image Search I selected three processes to apply which would break down the images from recognisable objects in to progressively more abstract fields of colour. Each process was applied to the original image in isolation rather than cumulatively. The original images were first converted to JPEGs and already being in the RGB colour mode, were given the Adobe 1998 ICC colour profile. They were then converted to Gifs so that they could be indexed using Colourphon, resulting in a 9 x 9 grid of dominant colours. The JPEGs were blurred with a specific and constant amount of Gaussian blur (a radius of 80 pixels) in Photoshop and finally, also using Photoshop, the RGB values for the pixels in each image were averaged to give a solid block of colour.

    4. Presentation of Results The methodology for the project resulted in a combined total of 1320 images, 120 for each of the basic colour terms. Each of the processes applied altered the coherence of the image to a different degree. They moved from defined forms, into indistinguishable shapes, allowing dominant areas of colour to appear more clearly and others subside, until finally, through averaging, a single, consolidated block of colour could emerge. As this project was a visual investigation, creating a body of images rather than generating pure numerical data or observations, the results were originally presented in book form. All of the images produced by each colour search term and the subsequent results of each digital manipulation were presented across sequential double page spreads. The book was divided into sections by colour term and although using the eleven basic colour terms it was not ordered in the sequence of colour term evolution proposed by Berlin & Kay as this order was not relevant to the project outcomes. While the book began with pink, the following sections were ordered spectrally: red, orange, yellow, green, blue, then by non-spectral purple, brown, black, grey and white. As the results were being presented visually it was felt that this was a suitable arrangement that would also be aesthetically pleasing to the viewer. The images were grouped by process, beginning with the original results and were presented in the sequence they ranked in Google Image Search results. These were followed by the groups of indexed images, the blurred images and the colour averaged images. This allowed the images to be compared both by hue and process and it was possible to observe how each effect altered the colour composition of the images.

  • 4

    The following images (figures 1-4) are a selection of examples from the book to illustrate the range of results produced by Google Image Search and by the subsequent digital manipulations.

    Fig. 1 Original images from Google Image Search for the colour term red, image results 1-16 as presented in their original book format.

    Fig. 2 Images from Google Image Search for the colour term red, image results 1-16, indexed using Colourphon, as presented in their original book format.

  • 5

    Fig. 3 Images from Google Image Search for the colour term red, image results 1-16, with Gaussian blur applied using photoshop, as presented in their original book format.

    Fig. 4 Images from Google Image Search for the colour term red, image results 1-16, RGB values of pixels averaged using photoshop, as presented in their original book format.

    Due to the nature of the project it is impossible to reproduce the results of the project in full for this publication. However if you wish to view the complete body of work it is available online and is hosted by Issuu at this address: http://issuu.com/Eleanorbydesign/docs/searching_for_the_rainbow/79.

    5. Discussion As a body of work this project presents a visual representation of each of the eleven basic colour terms, mediated by Google, and functions as a snapshot of colour on the

  • 6

    Internet that due to its constantly shifting nature, can never be replicated exactly. It shows not only the variety of responses to the names of colours but also the level of consensus across the range of images. It allows the reader to appreciate how the colours and images transform and mutate with each of the different processes applied, as they are transformed from clear shapes into a solid, uniform block of colour. As this study was produced as a visual research project, using design methodologies to investigate the relationship between colour and language there was no data analysis conducted on the results. However there is much scope for further extension of the project, particularly the potential for scientific or statistical analysis of the results. While this would clearly add more weight and academic rigour to the results of the project, at the time it was beyond the scope of my course. To have numerical data measured against accepted colour standards, supporting the visual output of the project may also benefit other researchers and make a valid contribution to the field of colour naming. However, one of the key aims at the outset of my main investigation into colour and language was to differentiate this project from other studies in the same field through the creation of a strong visual component to the outputs of the research. The project provides a bridge between purely artistic studies of colour and scientific investigations in the field, striking a balance between aesthetic experiments and hard data. While there are always areas for improvement in this respect I view the overall outcome as successful. 5.1. Extensions - further image manipulation As the use of Google Image Search proved to be a successful and relevant research tool I extended the project in two different directions while the wider investigation for my Major Project was still ongoing. In addition to the processes described in the methodology, I extended this line of enquiry by using a combination of the search results and the processed images of the basic colour terms. Working with the 1320 images already generated I considered methods which may generate further insight into the distribution of the colour relating to each term within the image results. All of the images for each process and within each colour term were layered with 30% transparency to demonstrate the combined effect of the colours in each image. It allowed the areas saturated with the colour used as the search term to be seen more clearly and to see how that particular colour used as the search term is distributed throughout the images. A sample of the centre area of the averaged images was taken and enlarged. This was one method for producing a culmination of the averages in the sample. It acted as a representation of that particular colour term, mediated both by Google and Photoshop. In addition to this, images have been created for each of the colour terms by taking a vertical, one pixel wide section through the layered images. This has been stretched

  • 7

    across a wider area to give a greater visual impression of the layering and to portray the colour composition of the image in an alternative way. Examples of the outcomes for this aspect of the project can be seen below in figures 5-10. The full results are available to view online at: http://issuu.com/Eleanorbydesign/docs/transforming_the_rainbow/11

    Fig. 5 (Left) Images from Google Image Search for the colour term red, layered with 30% transparency, as presented in their original book format. Fig. 6 (Right) 1 pixel wide cross section, taken from the centre of the image in fig.5 and extended horizontally, as presented in their original book format.

    Fig. 7 (Left) Images from Google Image Search for the colour term red, indexed using Colourphon then layered with 30% transparency, as presented in their original book format. Fig. 8 (Right) Images from Google Image Search for the colour term red, with Gaussian blur applied in Photoshop then layered with 30% transparency, as presented in their original book format.

  • 8

    Fig. 9 (Left) Images from Google Image Search for the colour term red, with RGB values average then layered with 30% transparency, as presented in their original book format. Fig. 10 (Right) Block colour created from the centre of fig. 9, an average of the average, as presented in the original book format.

    Collectively, these further visual experiments represented a conclusion to this line of investigation. While having created some striking and appealing imagery, the method of applying additional digital processes to the images presented a limited opportunity for further insight. However it is clear that the use of Google Image search, and other search engines, as research tools has great potential to generate insights, certainly in colour naming, if not beyond. 5.2. Extensions - widening the search Also building on the methods developed using Google Image Search I extended the process by widening the search to commonly known, but less widely used colour terms. Although this criteria is not as clearly defined as the eleven basic colour terms and therefore subject to the influence of personal bias, I felt it important to extend the process to include a greater variety of colour terms. The terms were all monolexemic, but many were descriptive, such as 'cream' and 'peach'. As before, the colour names were used as search terms in Google and the first thirty images from the ranking were extracted in the order they appeared in the results. The images were presented in book form with the colour terms ordered sequentially by hue, so it is possible to compare the results of different searches or similarly coloured images. In contrast to the original methodology the images were not subjected to any additional digital processes. This aspect study was specifically concerned with the comparison of Google Image search results, for colour terms that go beyond our basic rainbow. This method allowed for subtle distinctions to be observed between colours that are often regarded as interchangeable, for example wine, claret, maroon and burgundy.

  • 9

    It also allowed for similarities to be seen between colours that are generally thought of as more disparate. There was also a semantic dimension to the images, with some of the results for each colour term exhibited common themes. 'Champagne' was associated with wedding imagery, 'coral' was a strong fashion colour, and 'hazel' was almost exclusively linked to eye colour. A selection of the results generated by this part of the project can be seen below in figures 11-12. The complete body of images can be viewed online at: http://issuu.com/Eleanorbydesign/docs/looking_for_hue/21

    Fig. 11 Google Image Search results for the colour terms watermelon (left) and coral (right).

    Fig. 12 Google Image Search results for the colour terms aquamarine (left) and turquoise (right).

    The outcome of a Google Image Search may not provide a conclusive answer for the visual interpretation of colour terms, but as the results show, colour is rarely, if ever, definitive. Everyones idea of what a particular colour is, may equate to many subtly different shades. However, in all but a few cases, notably puce, there is some degree of consistency across the search results. While it is widely understood that both colour perception and colour naming are subjective, this methodology has produced a body of images defined by the process of colour naming but created in a real-

  • 10

    world, non-experimental context. The degree of variation seen within the colour range of the images is a reflection of how each colour terms is generally used.

    6. Conclusion Google Image search has allowed me to capture an interpretation of colour terms across a medium, which is still relatively democratic and where content is less mediated and regulated. The Internet comprises of a range of often competing influences from personal to commercial to political to academic, but perhaps the way colour is expressed in this virtual medium is actually is quite reflective of the practical ways we use colour terms for identification in the real world. The results of each search add to the collective understanding of what a colour term represents and it is possible for anyone and everyone, through their own contributions to the Internet, to influence that understanding in some way. This method is but one of the many possible ways of exploring the relationship between colour and language but it one that is truly a reflection of our digital age.

    Bibliography [1] Batchelor, D., 2000. Chromophobia. London: Reaktion. [2] Anon., 2013. W3Techs. Web Technology Surveys, [online] Available at:

    [Accessed 19/04/13] [3] Berlin, B. & Kay, P., 1969. Basic colour terms: their universaility and evolution. Stanford : Center

    for the Study of Language and Information. [4] Colourphon, 2008. [online] Available at: [Accessed 16/12/10].