• Skip to main content
  • Skip to primary sidebar
  • Skip to footer

Tuan Huang

  • Spring 2022
  • Fall 2021
  • About

Spring 2022

#final project: E-Forms

what about forms?

Since I came to America, I’ve filled out numerous forms, including VISA, I-20, W-2, DS-194,…etc. They all have their numbers and indexes. I realized that filling out forms means to apply for something, thus indicating I’m still an outsider for what I am applying for. Starting from the most basic part, VISA, just to be in this country legally, then I-20, a study permit, then DS-194, the record every time I depart and arrive at this country. And also lots of others forms that come after these, to deal with all aspects in my life here. I realized how few forms I had to fill out when I was living in my home country. Here, these forms are my identity; this is how the government has registered me, and thus, every time I travel, I have to bring a folder of testimonies. It feels inhuman when the officer admits the documents more than recognizing me as a person. And the languages used in the forms make me feel even more so – alienated and foreign. 

stuff I used

#01 Tracery

Starting in week 3, I used Tracery to create prompts for the forms. The results were sarcastic to me. The level of my incomprehension of the generated documents is almost the same as the forms I see in real life. The number of prepositions and nested nouns confuse me with the subject the question is asking. 

a list of computer generated form questions

#02 Transformer, GPT2

Using transformer and GPT2, I used the questions generated from Tracery as prompts, letting the computer answer the forms itself. I also added a government department on the top and a scary statement that its consequence always looks severe, but I also cannot not sign.

a list of form questions with different themes a list of form questions with different themes

#03 Standalone Python in Browser, HTML, Flask

Last but not least, to make those generated forms look official, I find the layout necessary. So I wrote a standalone python script in my text editor, adding HTML elements such as tables and hyperlinks, trying to build the structure of a form in the browser. Currently the HTML is written in a very messy way in Python, in the future I would love to integrate in a cleaner way to make it possible to use in a more generative way.

a screenshot of code in visual studio code

These are the main function code I wrote to run in a browser. As you can see the last part is so messy…(but it works!) Using the render_template enables me to write a simple CSS for the page.

return "<h2>"+dpt+"</h2><b>"+date+"<br><br>"+instruction+"<br><br><br><br>"+section+"</b><br><i>"+notes+"</i><br><br>"+"<table><tr><td>"+prompts[0]+"</td><td>"+output[0]+"</td></tr><td>"+prompts[1]+"</td><td>"+output[1]+"</td></tr><td>"+prompts[2]+"</td><td>"+output[2]+"</td></tr><td>"+prompts[3]+"</td><td>"+output[3]+"</td></tr><td>"+prompts[4]+"</td><td>"+output[4]+"</td></tr></table><br><br><p><b>"+vow+"<br><br>"+signature+"</b><u>"+computer+"</u><a href ='https://tuan-h.com'> [ print ] </a></p>"+render_template('index.html')

 

Final output

a screenshot of a browser with the form generator

 

Other thoughts & Future Development

  • use other corpora
    • instead of GPT2 -> for a more specific tone/ logic
    • choose corpus based on the government department?
    • the prompts are often too short to generate logical answers
  • include user input
  • instead of local host -> connect to the server and a domain

This is like an experiment with something that makes me feel alienated. Initially, I thought computer-generated content would be a good fit for it, that’s why I chose this direction. However, after spending time with it, I feel like there are a lot more aspects of this medium that are worth exploring. The authoritative stance, what questions to pose, what aspects the form wants to learn from its applicants, etc. By using different corpora and formats, I might be able to dive deeper into more specific themes and messages.

Link to Gist

 

 
 

 

 

#5 Form Generator + Auto-fill by computers

This week, we learned to generate text with pre-trained models and the usage of Transformer. I continued my form generator from assignment #3, which used Tracery to generate questions that we see on forms. 

little recap from the output last time:

lines of text that look like form questions

Later, the questions generated from Tracery get passed on to Transformer to serve as prompts, and the generated text becomes the answer to those form questions. 

Some examples of the output

lines of text that looks like form

 

lines of text that looks like a form

This ‘auto-fill’ content adds another layer of dissonance and alienation. The format of ‘form’ is almost unrecognizable without the hint of vow statement and signature. This form and its content become almost a sole creation of the computer (and ofc, I), which kind of delivers what I feel towards most of the bureaucratic forms and procedures. 

→→gist

#4 Revisiting Digital Cut-up

→ → gist

Using Spacy and Tracery, I revisited the digital cut-up I did in #2, and experimented more with what we’ve learned in the past few weeks. 

Inspired by Peter Handke’s poem “Song of a Childhood”, I started the experiment with the “When…” structure. This is the first rough draft, which one can tell the story by the key words and characters, even the paragraphs did not form a clear narrative.

screenshot with lines of poems

 

Then I tried to place the main characters (Lord, Adam, Eve, the serpent) in the main clause, followed by three clauses. In the end, I kept only Eve in the main clause, it started to look like another story told only from Eve’s perspective. This opens up another space to explore around –  the Garden of Eden before the incident happened, the Behind the Scene, etc. 

#1

  lines of poems

#2

lines of poems

#3

lines of poems

 

03: Form Generator

This week our prompt was to study a genre or a form of writing: link to GIST

The form I chose was “form”, as in forms of information or documents. It has little beauty in language and could scarcely be seen as literature. However, it indicates a lot of social contexts and points out the identity of an individual (or even, what symbols an individual and what not?) As an international student, I have been in numerous process of applications to move to America – Visa, bank accounts, Social Security Number, etc. These forms are filled with authoritative words that I could hardly understand, not to mention imagining the consequence of any violations. 

Some exemplars –

a form with input boxes and text a form with input boxes and text a form with input boxes and text a form with input boxes and text

At the beginning, I was trying to recreate a generator for those forms. To me, the random selection of words recreated the detached feelings I have for those forms. Skimming through the whole paper, usually I do not question the reason for the requests of my information, and skip through all the small texts and notes. And sometimes, the questions can be confusing too. For example, employed address, does that mean the address I will be working at, or the address of mine, the employed? 

Department of Immigration Services 
Country enrolled: _________ 
Hours per week estimated: _________ 
Current contact pursuing: _________ 
Preferred type: _________ 
Work country: _________ 
Designated type employed: _________ 
Hours per week of approval: _________ 
Country issued: _________ 
Status employed: _________ 
Current applicant: _________ 

I assure that, under penalty of perjury, that I have understand the statements. 
Signature _____________

Then I started to play around more by using ‘rules within rules’ that Allison mentioned last class, to make the whole thing even more alienating. 

Department of Career Development

	Second alien: _________
	Earliest document country: _________
	First work enrolled: _________
	Preferred position: _________
	Second approval: _________
	Registration that estimated document applicant approval number organization of alien that estimated date birth: _________
	Current organization of hours per week that enrolled hours per week of registration: _________
	Second official of program: _________
	Contact: _________
	First extension: _________

I assure that, under penalty of perjury, that I have understand the documents of void.

Signature _____________


04: Tunnel Cube

 

a white cube with led lights on a surface

Concept

Tunnel cube is an interactive project as an attempt to recreate the experience of cars driving through tunnels. When I was a kid, sitting in the backseat of my dad’s car, I was always fascinated by the lights and shadows cast on the tunnel because of the car lights. For this cube, the light follows the ball as it passes through the tunnel, and eventually comes out from the other side.

Process

In the cube, there is a slight ramp that makes the ball roll. At the end of the ramp, I attached a VL530X time of flight sensor to locate the ball while it rolls. And using this distance value, I map it to each individual Neopixel, which is attached to the outer plane of the box. Regarding fabrication, my biggest challenge comes from the linear direction of VL530X. Since it uses laser light for proximity sensing, the direction has to be very specific for it to work properly. 

a cardboard box with a ramp, a box, an Arduino board and a bread boad a cardboard box attached with led lights on one side

To make the transition of light smoother, I also used the average function in one of the examples from last week. 

a screenshot with lines of codes in Arduino IDE

a white cube with led lights on a surface

 

 

03: Text Performance

Love Is _

Concept & Structure

This week, our prompt was to do a text-based performance. With Valentine’s day around the corner, I was thinking about what love means to me. 

Love is…chaotic. That’s my conclusion. Love is chaotic, but mostly in a good way. It’s inseparable and connected. Therefore, I came up with a visual as an attempt to express this feeling. 

There are 3 layers in this piece. Starting with a rather clean style, as time proceeds, it becomes more and more chaotic and hard to recognize the words. Each word is composed of pixels of text. In the first layer, the dots are consist of “i”, “love”, “you”; the second layer, “i”, “hate”, “you”; and the last one, “i”, “hate”, “love”, and “you”. 

Making & Process

The process is an experiment with p5.js. In my sketch, I detected the brightness of each pixel and connected those values to my visual output, which were the locations and sizes of small texts. The computer was basically recognizing the brightest part in its screen. And to do so, I displayed and manipulated letter cards on my phone. 

a camera and a phone  a camera, a phone and a corner of laptop

Towards the end, I gradually lower the threshold for brightness, and that’s why the scene becomes messier and messier.

p5 sketch: click here

  • Page 1
  • Page 2
  • Page 3
  • Go to Next Page »

Primary Sidebar

Footer

Recent

  • #final project: E-Forms
  • #5 Form Generator + Auto-fill by computers
  • #4 Revisiting Digital Cut-up
  • 03: Form Generator
  • 04: Tunnel Cube

Tuan © 2025