I'm learning to code at ITP and when you're not fluent yet, it can get confusing. Oh hey! It’s just like my relationship with French, which I'm constantly trying to make sense of. I wondered what would happen if I used machine learning and predictive models to create actual nonsense French and then have a human try to make sense of it. To do this, I used Python, Markov chains, and Recurrent Neural Networks.

For the final presentation, I performed with artist Hugo Fortin. He's a native French/fluent English speaker and I'm a native English/elementary French speaker. I chose L'etranger as my source text for three reasons:

  1. The title means “the stranger” or “foreign”, which is how we felt reading the output.

  2. It’s the first text you’re given as a new French student.

  3. Camus’ style is very neutral with little decoration.

We ended up projecting our language biases onto the nonsense French the code created:

  1. Figurative meanings

  2. Sounds of words

  3. Exegesis or language interpretation

  4. False cognates

  5. Subconscious

Course: Reading & Writing Electronic Text

Brief: Use Python programming language as a tool for reading and writing digital text. Create and perform a piece of original text and present your methodology.

Role: concept, Python, performer

Source code: Python