Dragon NaturallySpeaking

What it is

Dragon NaturallySpeaking is a speech recognition software package developed by Dragon Systems of Newton, Massachusetts ^[1]. Its basic job is to turn speech into text on a computer, but it does more than dictation: the software can transcribe dictated speech, recognise and execute spoken commands, and read document text aloud ^[1].

Where it came from

Wikipedia lists the first release as June 1997 ^[1]. Dragon Systems released NaturallySpeaking 1.0 in 1997 as its first continuous dictation product ^[1], and BBC Chinese describes the 1997 product as the first continuous speech recognition product ^[2].

"Continuous" is the key word. It means the software follows speech spoken in a natural flow, rather than requiring the speaker to break every sentence into separate, isolated words. Peter Mahoney, whom BBC Chinese identifies as a senior vice president at Nuance and general manager of Dragon, said Dragon pioneered continuous speech recognition and made practical dictation possible ^[2].

How it works

When someone dictates, the words the software has recognised appear in a floating tooltip; they are transcribed into the active window once the speaker pauses ^[1]. That design lets a user watch the recognition happen and see a sentence completed before it lands in the document. The same program also listens for spoken commands and can read document text back aloud ^[1].

Speed was part of the pitch from the beginning: BBC Chinese reports that Dragon NaturallySpeaking could recognise speech at 100 words per minute ^[2], which is roughly the pace of ordinary conversation rather than the halting delivery older dictation tools demanded.

Later versions and ownership

The product has changed hands several times. The product line passed through Lernout & Hauspie, ScanSoft and Nuance Communications, and later to Microsoft ^[1]. Microsoft announced plans to acquire Nuance in 2021 and completed the acquisition in March 2022 ^[1]. Wikipedia lists version 16, released on 28 February 2023, as the stable release ^[1].

Its place in the field

The software's lasting niche has been professional paperwork. BBC Chinese notes that doctors in the United States and United Kingdom still used it to file medical records at the time of its report ^[2] — a use case that suits dictation well, because a clinician can speak findings faster than typing them, and the vocabulary of a speciality repeats predictably.

Seen from a distance, Dragon NaturallySpeaking is a useful marker of how far consumer speech recognition has come: a product line that began in 1997 with a continuous dictation program was still being released in version 16 in 2023, having passed through four corporate owners along the way ^[1]. It also shows how a research-heavy technology finds its market — not as a general-purpose novelty, but as a tool for people whose work is mostly words.