Bryan Ma


Definitions is a new media installation representing the culmination of my MFA thesis research at Parsons. It utilizes the ability of computers to comprehend semantic meaning via "common-sense networks" to poetically represent the socio-cultural effects of natural language processing.

The installation is made up of 15 small LCD screens that individually display seemingly random English words or phrases one at a time in sequence from the left to the right. When all screens have filled up with words, it restarts and displays a new sequence.

Each LCD's word or phrase is connected via some semantic relationship to those adjacent to it - for example: "computer" -> "keyboard" -> "music"; or "cat" -> "water" -> "plant" (computers [have] keyboards and keyboards [play] music, and cats [don't like] water which is [required by] plants). These words and phrases are sourced in real time via an algorithm searching through MIT's massive ConceptNet semantic network, so the combinations are effectively endless, and occasionally highly amusing or unexpected - though with a degree of scrutiny it becomes apparent that the first and last LCDs always have the same two words: "person" on the first, and "money" on the last. The semantic pathway between them, though innumerably varied, begins and ends identically each time.

With Definitions I hoped to poetically model the role of computational language recognition networks in the commodification of previously uncommodifiable human data - as we give computers better ways to understand us, we also enable others to capitalize.

Social networks commodify social interaction with keyword recognition among many other techniques by selling user data for targeted advertising, and human-generated search data on Google becomes part of the product itself by wrapping up its query and results as part of the search engine's driving algorithm - as does the increasing ability of computers to understand language seem to suggest that a fundamentally intangible aspect of human experience has the capacity to be filed, sorted, rationalized, manipulated, or commoditized. This is not to discount the humanistic goals that such technological progress has the capacity to fulfill, but predominant societal paradigms (in our case, that of globalized capitalism) remain capable of co-opting such gains to their own ends.

I developed Definitions primarily in Processing and Arduino, writing a recursive depth-first graph search algorithm to find connections between my target nodes using the ConceptNet API. As searching through 15 levels of connections is not computationally trivial, it required some tricks to keep a diverse set of results coming through, which were then archived and displayed via serial across the LCDs, driven by a single Arduino Mega.