Comparison with automated systems

You may be tempted to use an automatic speech recognition ("ASR") system to complete your dictation. However, these systems will struggle with the normal style of dictation of most authors. We know from experience that many authors are in a rush and are dictating in imperfect, noisy, conditions or are using technical terms of art or proper names which the ASR system will be unfamiliar with.

It is often the case that a document processed via ASR will then need a thorough human review to ensure that all the dictation has been heard correctly and that spellings are correct and consistent. Where you have a "batch" system which processes the dictation in the background, someone will often have to relisten to the entire dictation to check that the correct text has been transcribed by the automated system. Where you use a "streaming" system (where the transcription is done live) then you have to check as you go along that the right thing has been transcribed, which will often interrupt your flow and mean that a document takes longer to produce than it should.

At Indraft, we use ASR technology to produce a draft, but we have a team of human proofreaders to produce the final transcript that we return to our clients. This results in a much higher-quality product than using technology alone.

Advantages of human transcription

An ASR system is liable to make more mistakes than humans. This is especially the case with regard to company names, people names, and with homophones: this means you spend more time correcting the transcripts.
An ASR system will not generally apply the correct formatting to a document – bullet points, numbering, bold and italics are just some examples where an ASR system will not produce the best result. Where you are dictating a letter, you will have a standard template with spaces for references, the letter heading, the address, etc. Without significant interaction/post-processing amendment, the result of a pure ASR system will be inadequate.
Humans can cope with a wide range of templates\document formats from being given a voice instruction. An ASR system will often just produce a blank document, or you spend time opening the right template and moving the cursor around (for streaming transcription) or later copy-pasting the ASR transcript to the right template (for batch transcription).
Humans can do template amendment work that an ASR system cannot cope with. Detailed templates are a mainstay of many professions, allowing standard wording to be included where necessary, as well as prompting the author with the points to include in a particular report. You may have a firm-wide template which you know contains the right wording but that has several options in it – processing the necessary changes using an ASR system is not possible. You need a human to listen to the dictation and make the necessary changes to the template. Sending all your document dictations to Indraft means you don't have to worry about whether they require using a detailed template or not: we can work with your standard documents just as your in-house support staff would.
When you are dictating, you may often want to produce several documents at once – a reply to the other side, follow-up questions to the client, an update to the agents, send further queries to a management company, etc. Within your dictation, you may be jumping around these various documents. An ASR system is likely to create one document effectively combining all your separate correspondence, requiring someone to go through and check what is meant to be where. In contrast to this, a human transcriptionist will produce all the required documents and put the dictation into the correct document.
If you have a travelling draft document that is being amended by tracked changes, often it will be easier to work with that document using a human transcriptionist rather than with an ASR system. You can jump between pages easily, you can add Word comments, and you can make additional changes or accept or reject existing changes.
Small mistakes which are more likely to be made by an ASR system could have crucial consequences in a legal or other professional advice context. It is important to have your documents reviewed by a human for sense checking and to ensure that the correct context has been grasped from the dictation.