|
Getting your Trinity Audio player ready...
|
If you’ve worked in media, you know what it means to be overstretched. Reporters, producers and editors are hostages to time. There are a lot of forces at work these days that make it even tougher than it used to be: everyone is under increased pressure to provide insightful content in the blink of an eye. One of the most frustrating-but-necessary parts of the process is the oh-so-tedious task of transcribing interviews and news conferences – anyone who has transcribed knows that it’s both time-consuming and inefficient.
Traditionally, people working in media have transcribed audio recordings manually. But the slow pace and high cost of manual transcription work against them, which means writers and editors are struggling to keep up with the fast pace of today’s content churn.
Over the last ten years we’ve seen a huge shift in the way we consume news and content. Not surprisingly, more people than ever are using the Internet, which means a lot of people are now reading or downloading online news, newspapers or magazines – a jump from 20 to 64 per cent in just ten years. Because of this, journalists are frantically trying to produce content quickly and efficiently to meet this growing demand.
And there’s more. The way consumers interact with content is changing. Audio and video are growing in popularity – with 65 per cent of Internet users regularly viewing video as a core online activity. It’s clear that digital publishers need to change to keep up with consumers. But how can they be sure they’re giving consumers what they want, when they want it and how they want it?
Combining transcription with technology
As with most new technology, efficiency and ease of use are the most important things when it comes to digitising the transcription process. The last decade has seen impressive new developments in machine learning and speech recognition that have revolutionised the way audio is processed – giving rise to a new wave of automated transcription services.
One of the areas that’s seen the most progress in recent years is automatic speech recognition (ASR). This technology uses artificial intelligence (AI) to interpret the sounds that make up human speech and convert them to text. Using this technology, publishers can transcribe audio files at a fraction of the time it takes a human, helping to tackle a once complex and inefficient workflow.
Instead of spending an average of six hours a week transcribing audio, slowing down the reporting process and newsroom productivity, journalists now have access to transcribed audio and video in minutes instead of days. In addition, AI transcription makes content searchable, editing easier and the entire process more streamlined.
The accuracy challenge
Producing accurate transcripts is essential for people working in digital publishing. Interviews have to be recorded accurately and clearly for reference and quotation. As a result, many choose to pay for traditional manual transcription, which boasts up to a 99 per cent accuracy rate. However, what many fail to realise is that further costs are often added for this level of accuracy.
While a lot of things can affect the accuracy of machine-generated transcripts, including background noise and multiple speakers talking over each other, users can minimise these to greatly improve the accuracy of digital transcription. With clear audio, AI transcription can be as accurate as manual – with a recent index of UK politicians showcasing levels as high as 98 per cent. And while AI transcription is not yet completely error-free, it only takes a fraction of the time as a human transcriber, which in turn frees up the user’s time to make quick corrections.
To make the editing process more efficient, text in the transcript is stitched to the corresponding moment in the audio or video using AI, creating interactive transcripts that make it easy to find important moments or locate keywords – not only saving time but helping to ensure accuracy. Alongside this, custom dictionaries can help AI technology recognise certain words, like brand names and technical terms, to help further increase accuracy.
There’s also the issue of security: the more eyes and ears that have access to your content, the less secure the data becomes. Manually transcribed files are often divided into smaller files and sent to multiple transcribers; once transcribed, an additional person needs to put the different parts of the transcript back into a single document; and yet another person will proofread the transcript before returning it to the client. That’s a lot of eyes and ears being exposed to potentially sensitive information – the opposite of secure.
Letting automated transcription do the heavy lifting and giving users the opportunity to then search and verify provides a new, efficient solution for both accuracy and speed. AI transcription produces a first draft for smooth and fast editing. Then, using the time saved from manually transcribing, the user can edit the text for 100 per cent accuracy. As more publishers use automated transcription, the technology will continue to evolve and speech-to-text algorithms will continue to improve in both accuracy and inefficiency. And with machine transcription, no one sees your data but you, keeping it secure.
Digital advances like automated transcription put the power of time back into the hands of the user, helping to relieve some of the pressure that media professionals experience so they can create content quickly. Now, instead of being weighed down by transcription, people in digital publishing can spend the extra time doing what they do best: producing creative and compelling work.
Jeff Kofman, CEO and Founder, Trint
About Trint:
Trint is the first automated transcription service to combine a text editor and an audio/video player into one easy-to-use tool, making it possible to search recorded content by keywords, verify by listening to just that portion of an interview and edit a transcript to perfection in minutes.
Founded in December 2014,
Photo by Leon Contreras on Unsplash
