interviews

Verbatim transcription: chasing the ideal

Transforming speech into text is not as simple as it may first appear. Professor Michael Toolan, Head of English at the University of Birmingham, discusses the linguistic issues surrounding digital dictation.

The differences between writing and speech may seem minimal, but their characteristics entail a disparity which can hinder transcription.

Speech is a continuous vehicle of expression, with far less premeditation. As such, the level of technical accuracy in our language is much higher in writing, when we consider our words and constructions before we commit them to the page.

Medical transcription is no exception. Professor Michael Toolan, head of English at the University of Birmingham, talked NHE through some of these obstacles.

Ambiguity

Ambiguity in speech, through words or their structure, becomes more evident when written down. Without explicit directions, which can require a modification of natural speech, this can result in inaccuracies or misunderstandings.

Toolan said: “Some of the ambiguity could be lexical. For example if a doctor says into her dictaphone: ‘I saw Mrs Smith yesterday and she seemed quite unwell’, a word like ‘quite’ could be rather ambiguous, as to whether it means ‘very’ or ‘only a little bit’.

“That’s something the physicians have to become adept at, speaking in a way that’s going to look okay in writing. I think a lot of people do have that; they develop a dictation style of more formal speech, because they’re really instructing someone or something to write out their words later.”

Secretaries & segmentation

Toolan suggested that an understanding of medical jargon would help secretaries to transcribe dictations accurately, although categorising the flow of speech into discrete words was a challenge for any type of transcription.

He explained: “Even if transcription is being done automatically by voice recognition, presumably somebody goes and checks it afterwards and cleans it up. So whoever that person is, they should be well versed in medical terms and won’t be thrown like a word by ‘hypoglycemic’. They’ll recognise it; they won’t be tempted to split it into two words.

“Certainly the whole business of segmentation of the flow of speech is always a problem in any kind of dictation exercise.

“But people who get familiar with a sub-language, like medical discourse, will get quite used to certain set phrases. And I suspect there is quite a lot of formulaicity about these dictated reports.”

Grappling with challenges

On overcoming these challenges, Toolan suggested important factors include using the best equipment possible and choosing location at the point of dictation with consideration for acoustics and background noise. “All of these simple things could make the product of the dictation harder or easier for whoever it is subsequently to decode,” he said.

Additionally, a transcriber of a significantly different accent group to the doctor could be challenging to interpret accurately.

Even heavily pronounced regional accents could present comprehension difficulties.

Toolan said: “If the person ‘cleaning up’ the dictation doesn’t have some familiarity with that doctor’s accent, that could be a problem.

“These are things we frequently have to grapple with in interaction, but it’s particularly critical in the medical field.”

No guarantees

When a lack of accuracy could have direct negative effect on patients, the content of the transcription becomes particularly important, but as Toolan commented: “There are no guarantees with language communication.

“We hope for a certain degree of stability and that’s why we have dictionaries and a practice of widespread agreement on the meanings of words.

“Even with all of that, there is no absolute certainty that when I say something you will get exactly what I had in mind. In fact that may be an unrealistic goal to aim at. You do rely on a high degree of convergence when speaking.

“One of the ways language grapples with this,” Toolan continued, “is by means of redundancy.”

He suggested that redundancy, or repetition, is one of the key ways in which we ensure the most critical information is portrayed.

“It’s a form of checking, with important instructions repeated. I suspect that doctors almost spontaneously repeat the crucial things they want to say. That this patient needs a certain medication, those kinds of things might get repeated just to be safe.”

Every word

Transcribing every word would mean that every mispronunciation, every false start and mumble would appear as text. This could require more editing, but allowing a level of interpretation for transcribers gives them the chance to misjudge importance when deciding what to include in the text version.

“Should they transcribe every word? In some situations it might be regarded as very unprofessional and even unethical not to transcribe every word. The whole business of verbatim transcription is a bit of an ideal.” 

Tell us what you think – have your say below, or email us directly at [email protected]