# Speech Recognition **Domain:** Natural Language Processing / Signal Processing / Artificial Intelligence **Doc Type:** Discipline Node **Maturity:** Developed **Related:** [[Natural Language Processing]], [[Shoebox]], [[Machine Mediation]], [[IBM Research]] ## Definition **Speech recognition** converts acoustic speech signals into words, commands, or other machine-readable representations. It must bridge continuous, variable sound and discrete linguistic structure. ## Early IBM Waypoint IBM engineer William C. Dersch introduced [[Shoebox]] in 1961. The experimental system recognized spoken digits and a small command vocabulary, then directed an adding machine to perform arithmetic. Its competence was narrow, but its public effect was significant: voice could become an input to computation. ## Corpus Function Speech recognition is a form of [[Machine Mediation]] that turns the body’s ordinary vocal action into control infrastructure. It extends agency when it increases access; it can also become a layer of [[Invisible Mediation]] when capture, transcription, or interpretation occurs without meaningful consent. ## Sources / Provenance - IBM, “Speech recognition”: https://www.ibm.com/history/voice-recognition