Welcome to the home of WikiTADA.VizTrans. This is a web-based collaboration area for ...

viztrans_logo1.jpg

"Showing Text in Image Format" - Jamie Demeter

VizTrans ver.2DO3c is ONLINE. Please feel free to try it out here.

Table of Contents

About the Project

What is VizTrans

The name VizTrans is short for Visual Translator.

VizTrans is an experimental icon based text visualization tool. Due to the inherent nature of icons, the processed texts become extremely simplified, and can be difficult to understand if one is unfamiliar with the social environment in which the icons were selected. However, this leads to an interesting analysis of the contextual environment of the processed text, in addition to the processed text itself.

Our goal is to create a tool which can accept texts from various input methods, and can then apply a broad database of concepts to iconify the input text. However, due to the constraint of time, we are limited in the number of icons we can create. As a prototype our team has chosen eight scripts from various movie genres, providing a range of theme and content while reducing the number of concepts that need to be iconified.

Language V.S. Text: What’s the point?

How does it work?

VizTrans parses a submitted text and replaces words with HTML image tags based on a predefined list of words/icon groups. There is an option to append the list with words/icons for the actors and characters in the sample scripts. The processed text is then displayed below in the output frame. Currently there is only one mode of processing, and the processed text is displayed with both unreplaced words and the icons that replaced words.

In the future, we hope to include more display modes, and also have the option to switch between them quickly. Here is a list of some of the display modes considered:

Alternately, instead of using a predefined set of words and icons, we considered creating a protoype that used a thesaurus to simplify the text and then referenced Google image search and adapted the most relevant image for each word to use as the icon. However, since the results returned by the thesaurus and Google image search are generally unreliable, this kind of program would more of an interesting toy rather than a useable text analysis tool.

How VizTrans was made (Development Overview)

During the development of VizTrans, the developers were divided into two groups: The Icon team, responsible for word selection and icon creation, and the Programming Team, responsible for creating the VizTrans script and interface.

Word Selection and Icon Creation

Having a complete list words and a library of clear icons is crucial to the functionality of VizTrans. If a word is missing from the list, VizTrans is unable to identify and replace it. If the icons are hard to recognize or are not immediately obvious as to what they represent, then the output from VizTrans is not useful. However, building a "dictionary" of words and icons comprehensive enough to allow VizTrans to deal with any English text is a daunting task, and would likely require more time than we had available to us during our class.

In order to scale the project down to something feasible within our time constraints, we decided to limit the words in the dictionary for the 2DO3 version of VizTrans. Ideally, VizTrans' dictionary would include every word in a modern English dictionary, including all adjectives, adverbs, conjunctions, determiners, interjections, nouns, prepositions, pronouns, and verbs. However, since one of the goals of VizTrans is to streamline reading, we agreed that it would be acceptable to simplify our dictionary by grouping similar words of a type, and even eliminating some words or types all together. It was determined that the number of nouns, verbs, adjectives and adverbs could be reduced by grouping similar words, and interjections, and some determiners and conjuctions could be omitted without losing much essential meaning.

To further limit the size of our dictionary for this project, we chose to include only the words used in a selection of movie screenplays. However, since we wanted to include a direct form input option in VizTrans, to ensure there was still an interesting range of words in our dictionary we chose at least one screenplay from every major genre of film.

To effectively address the critical nature of generating a strong dictionary, we formed a subgroup called the Icon Team to focus entirely on word selection and icon creation, and most of the VizTrans group members were assigned to this subgroup. The Icon Team was responsibile for analysing the screenplays and selecting the words that would be replaced with icons by VizTrans. To do this, each screenplay was first run through Taporware (http://strange.mcmaster.ca/~taporware/htmlTools/listword.shtml). The tool returned a list of which words were used in the screenplays and how many times each word appeared. After examining this list, the Icon Team had to decide which words were similar enough to be grouped.

Having decided on the words that would be recognized by VizTrans, the Icon Team had to come up with icons to represent each of words and word groups. We decided to make icons with the dimensions of 512x512 pixels since they could then later be easily resized without difficulty, but this meant that it was more difficult to find appropriate source images. Once a set of icons were selected, they were reviewed for effectivness by the whole group, and any rejected icons were redone.

Finally, a list of word groups was created in YAML markup format, to be used by VizTrans.

Program Creation

The programming aspect of this project was more learning than it was actually writing code. Of the three members in the Programming Team, only one actually had much of a programming background. The Programming Team was responsible for two important parts of VizTrans: the word replacement script, and the graphical user interface.

In terms of the word replacement script,

Coming up with the idea

The idea originated in class when topics had to be chosen for our next project. Isaque Hossain and Kyle Katsuyama were the developers who wanted to see if it was possible to use images as words. While image symbols are all around us, most of them have words within the symbol to make sure they can be understood. For example the road sign "Stop" has the word stop in it. It is clear what that symbol is representing. But would the symbol maintain its original meaning if the word was removed? That is what started the project. However, because this is an experimental prototype only a limited amount of text would be used for the project. It was decided it should be limited to eight movie scripts covering different genres. This process was not random as scripts were chosen for films which redifined their particular genres.

Choosing Scripts and Words

The very first part of the project was to get a list of text that we would use for the experiement. We wanted the texts to be similar so there could be some kind of connection between them, but the texts also needed to be different. Thus, movie scripts of different genres were chosen. Once we had our eight scripts the next phase of the project was to know what words we should make into icons, what words we could remove, and what words we could combine with others. To do this we ran the eight scripts through Taporware: http://strange.mcmaster.ca/~taporware/htmlTools/listword.shtml. The tool gave us what independant words were used and how many times they were used. We then listed the words in a excel file and decided within the Icon team which words could be combined. For instance no and not were combined as well as heart and love.

Making the Icons: Do you get the word?

We made the icons with the dimensions of 512x512. At this point in time we did not know how large we would show the icons so we wanted them pretty large so if we had to make them smaller scaling them down would not be an issue. Once we had our master list from the excel file our team began to make the icons, seperating them into pages. This was not difficult to do for the most part. How else are you going to represent the word run without showing something running, so most icons were pretty straight forward. However, some words we struggled with, such as: main, across, and close, are just to name a few. We were avoiding puns because the user may believe the word "main" to not be "main." In one instance, when we were using puns we represented the word "main" as a sewer meighn. However, the user might not know what a meighn is, or if they do they may not understand why it is being shown. There was a consensus finally reached among the icon team. Yet, this was not the end of our problems.

Images and their meanings...their many, many meanings...

After we had created our icons, we showed them to the rest of the team. Many of the icons did not capture the correct meaning for certain words while others could not be understood at all. We then had to talk about what each word meant to us and how could that word be best represented. The word "across" may seem easy to do in an image, but it proved harder than we thought. Also we had to discuss how we would represent words such as "life", and of course, "main." We decided if the user could get the idea of the word, or grasp its meaning, then we would have a better chance at having the words make sense. Based on our discussion the icon team went back and re-made, most of the icons with what the team had discussed in previous meetings. We then listed the words in a text file according to how the programming team told us to do so. While we were making the word icons we also had to make a seperate file for celebrity icons. These were the characters who appeared in the script. We also decided to obtain other characters and people who did not appear in the script. This was done so the user would have the option of changing the characters around in the script to see how much the meaning of the text was changed. This was not a difficult task proving to be very straight forward. As for the actor selection, it was very straight forward. Aside from those whom appear within our selected scripts, a whole variety of famous actors were included so that the user will be able to select different celebrities to act as a certain character of a script.

Researching

Since our project was about converting words into images it was necessary to study similar techniques used by other scholars. We wanted to understand some of the ramifications of making something like this. Would the user understand the word? If not, would the user understand the symbol as a different word and would that change the meaning of the whole structure?

Learning the programming

The programming aspect of the project was a combination of the object oriented language of Ruby and Flash's Actionscript. The user interface is designed with Flash, the various navigational properties and settings of the program is/will be coded with Actionscript. The main search and replace algorithm is written in Ruby. The code reads the input text and the search/replace keywords from a file, and by using regular expressions, replaces the keyword with an url pointing to an image file to be replaced with. Ruby was used, quite frankly, because that was the program which was best suited towards our needs. It had all the neccessary components for us to create VizTrans.

Communicating

Communicating with the different team-members was, at first, used through e-mail. While this technology provided very interesting possibilties for communication it was limited. Therefore, the move towards using a Wiki page was a natural step. The Wiki allowed users to not only edit and change information, but also allowed the different users to see the changes that were made. In addition to this we met every Wednesdays at 11:30am and discussed what we had done and what still needed to be done. Further talks took place on the Wiki and during our usual class time.

Combining

Finally, we have to bring together the work of the Programming and Icon teams. Thanks to the modularity of the project, this was actually quite simple. The finalized icons and word lists created by the Icon team replaced the temporary icons and word lists used by the Programming team during development, and if the Programming team were successful everything should work properly.

The Cutting Room Floor

Due to this project being an experiement it is only natural that some things that were created and imagined were left out. In one example a website was constructed featuring an interface, logo, and updated information. What became apparent to us was that we already had a website. The VizTrans Wiki was it. All of our information was already on the Wiki and anything else added to the website would be the same as the Wiki. Because of this we opted to not create a website, instead providing a link: Interface Design, to the screen shots of what the website would look like if we had more time to create VizTrans the way it was first imagined. We also did create another alternate interface yet the one above was chosen to be the one we would use. Where it stands VizTrans is a prototype experiment in which the user sees specific images which represent the corresponding text.

Academic Study and Research

Searle, John. "Minds, Brains, and Programs."

This article be found online at: http://www.bbsonline.org/documents/a/00/00/04/84/bbs00000484-00/bbs.searle2.html

Srihaki, R.K. "Combining text and image information in content-based retrieval."

Yildiz, Melda "Semiotics of New Media Literacy."

For more information on this research, please visit: http://euphrates.wpunj.edu/faculty/yildizm/SP/#Research.

Project Developers

During the course of 2DO3 in 2005, the members of the "Thousand Words, One Icon" Comittee were:

You can view our personal opinions and comments about the project here

For the Developers

Latest Updates

*Almost Finished! The project is being handed in tomorrow December 7th 2005. Celebrations are in order -December 6th 2005

Teams and Section Pages

Attachment sort Action Size Date Who Comment
optimis.mid manage 20.2 K 16 Nov 2005 - 02:29 JamieDemeter  
sinclairdance10.gif manage 40.3 K 16 Nov 2005 - 16:36 JamieDemeter  
viztrans_logo1.jpg manage 11.9 K 20 Nov 2005 - 09:29 IsaqueHossain  
images_actors.rar manage 10457.0 K 11 Dec 2005 - 19:12 GaryCheng