Spokentowritten

Spokentowritten is the practice and study of converting spoken language into written text. It encompasses both manual transcription by humans and automatic transcription produced by software, and it involves decisions about orthography, punctuation, capitalization, and annotation of speakers and other non-speech events. The term captures the full process of rendering oral discourse in a written medium, with attention to how faithfully spoken form is represented and how the result will be used.

In practice, spokentowritten operates along two main tracks. Manual transcription relies on trained transcribers who listen to audio and produce transcripts, often following specified style guides and transcription conventions. Automatic speech recognition technologies convert audio directly into text, typically with post-processing steps to improve accuracy and align transcripts with time stamps or speaker labels. Transcripts themselves come in various styles, from verbatim records that preserve fillers and repairs to cleaned transcripts that remove disfluencies and standardize spelling and punctuation, and from speaker-attributed lines to multi-speaker dialogue formats.

Standards and tools in spokentowritten include transcription conventions, time alignment formats, and encoding schemes such as those used in the Text Encoding Initiative. Outputs may take plain text, subtitle files, or richly annotated formats used in linguistics and digital publishing. Evaluation metrics like word error rate gauge automatic systems, while human review assesses readability and usefulness.

Applications span accessibility (captions for media and assistive technologies), journalism and media production, legal and court reporting, education, and language documentation. Challenges include handling dialects and noise, representing non-speech sounds, and balancing fidelity with readability and privacy concerns. Related topics include automatic speech recognition, transcription standards, and speech annotation.