The Rectification Train: Recovering Meaning from Everyday Typing Errors Without Altering the Record

Abstract

People type faster than they type accurately, and a growing share of the record, from chat logs and working sessions with AI systems to field notes and first drafts, is captured exactly as typed. Uncorrected text is often unreadable to search: the words are there, but no query can find them. This paper describes the rectification train, a staged method for recovering the meaning of damaged typed text while preserving the original exactly. It sets out why typing errors take the shapes they do, drawing on four decades of research on skilled typing as motor control; presents an eight-class taxonomy of typing anomalies, six classes from that research and two named here, including the Perimeter Key Co-activation Destruction Event (PKCDE); describes the six processing stages and the rules that govern them; records how the method was developed between 27 and 30 May 2026; and reports a computed test on thirty-one turns of everyday, uncorrected typing. In that record, search returned at least one hit for 8 of 15 fixed questions from the raw text and for 15 of 15 after rectification: a 1.9× improvement by question and 2.5× by turn. The method is a technique of Latent Value Query and runs on Generative Temporal Referencing. This paper discloses the concept, taxonomy, stages, and results; it withholds the implementation. It has not been peer reviewed.