This body of work addresses the chilling scale and effect of state-enacted censorship, enforced through digital infrastructures. The three works presented act as lyrical repositories for suppressed voices, with each technically scrutinizing and poetically portraying tweets censored from Weibo, one of the biggest social media platforms in China. The Unerasable Series explores the politics of erasure and the temporality of voices within the context of digital authoritarianism. It presents the sheer scale of unheard voices by technically examining and culturally reflecting the endlessness, and its wider consequences, of censorship that is implemented through technological platforms and infrastructure. 

The series of Unerasable Characters collects unheard voices in the form of censored/erased (permission denied) textual data. This is based on one of the biggest social media platforms in China—Weibo—via the system called “Weiboscope,“ a data collection and visualization project developed by Dr. King-wa Fu from the University of Hong Kong. The system has been regularly sampling timelines of a set of selected Chinese microbloggers who have more than 1,000 followers or whose posts are frequently censored. 

Unerasable Characters I presents two components in relation to Machine Learning: First is the input data that was collected between 30 June 2021 and 30 June 2022, containing 54,064 sample censored posts and presented in a stack of papers with more than 6,000 pages. Second is the predictive/output data from each iteration of machine learning training processes, based on the censored input text, in the form of a DIY book with customized binding tools. Inspired by the community movements especially white paper protest and blank paper revolution, the book is meant to be generative yet unreadable to circumvent censorship. This also challenges the common use of machine learning with precise accuracy and effective production. The project asks: is this a forbidden book? 

Unerasable Characters II consists of a custom-software that scrapes the erased “tweets” from Weiboscope on a daily basis, and the project presents the living archives in a grid format. Each tweet is deconstructed into a character-by-character display occupying a flashing unit for a limited period. The duration of the visibility of each ‘tweet’ is computed from the actual time the post was present on Weibo before being removed from the platform, transforming from a busy cacophony of voices and characters to a silent and empty space, marked by the disappearance of the characters. New archives which populate the work continue to be retrieved endlessly. 

Unerasable Characters III utilizes data between 1 December 2019 and 27 February 2020, the time when the COVID-19 outbreak was started in China. According to King-wa Fu & Yuner Zhu, there were 11,362,502 posts during the period, among which 1,230, 353 contain at least an outbreak-related keyword and 2,104 (1.7 per 1,000) posts had been censored. The artwork displays all the erased archives in the format of a web presentation, where each tweet is unreadable. The content has either been obscured or blacked out, except the punctuation​, emojis, and special characters. However, what remain are the pauses and blurry timestamps, depicting the affective and expressive, as well as temporal and spatial dimensions of unheard voices. Users can interact with the web by pointing to those pauses, contemplating the poetics of silence and erasure, and further questioning how the culture is being normalized via systematic processes and political infrastructure.