When mechanical Chinese typewriters first entered the marketplace in the 1910s, they featured tray beds containing approximately 2450 free-floating metal characters arranged within a rectangular matrix. Characters were arranged according to the Qing dynasty reference, the Kangxi Dictionary, whose “radical-stroke” system had for centuries formed the basis of an immense and highly diverse information infrastructure encompassing dictionaries, indexes, catalogs, name lists, telegraph codes, typewriters, and more. Beginning in the Republican period (1911-1949), linguists and engineers experimented with alternate organization and retrieval systems, witnessing a proliferation of competing taxonomic systems. It was not until the early Communist period (1949-present), however, that a decentralized network of largely anonymous typists broke with tradition and began to develop natural-language systems of categorizing Chinese characters. Rather than following the radical-stroke system, they sought to maximize the proximity – if not adjacency – of those characters that, when paired together, formed the most commonly used two-character compounds (known in Chinese as ci) and political phrases. What these typists conceptualized and created, this paper will argue, was a technology of probabilistically anticipating the instantly immediate future - or, in other words, the conceptual and practical foundations of what is now referred to as “predictive text” or “autocompletion.” This paper will examine the epistemological, technological, and sociopolitical foundations of this technology, one that has become central to modern computing, commerce, and governmentality.
Image
