Leading publishers including PBS and the San Francisco Chronicle are utilising AI-driven speech-to-text technology to streamline their production and turn words into text.
Speech to text platform Trint has been developed by former CBS news correspondent Jeff Kofman as a productivity tool for journalists inspired by his own frustrations juggling multiple tasks at speed.
Trint uses artificial intelligence to enable users to transcribe, edit, translate and share files easily.
This allows them to quickly search transcripts, pull out key quotes, generate closed captions, and collaborate with colleagues to share stories.
The technology is already being used by media groups including The Washington Post, Der Spiegel and the Center for Investigative Reporting.
Helping editors quickly find broadcast snippets
The first is to identify key sections of interviews for use in broadcasts, something that would have previously been an arduous task. A broadcast interview might take 20 or 30 minutes, with only a minute of sounds actually broadcast. Trint helps editors to quickly find snippets.
Creating transcripts of the audio enables the NewsHour team to search through the copy and pinpoint exactly what’s needed for different packages.
Maintaining quality, despite the pressure to be first with breaking news, is also a prerequisite. Trint transcripts come back with well over 90% accuracy.
PBS also uses Trint for a secondary purpose: interviewing potential guests for its evening show to get a sense of what they might say in the discussion.
Putting the interview on the platform enables the audio to be swiftly converted into text. This transcript can then be forwarded for the approval of senior editors.
Finding ways to reduce the time spent manually transcribing interviews and speeches was the original inspiration for Trint founder Kofman.
The experienced broadcaster, whose 30-year career included stints as a foreign correspondent, had seen reporters coming under intense pressure.
“The demands are overwhelming,” he said. “I saw the opportunity to build a productivity tool that could make the workflow smoother, easier, faster.”
Kofman teamed up with developers to explore how the increasingly accurate output of artificial intelligence could ease the burden of transcription.
How Trint voice-to-text platform works
Artificial intelligence is at the core of the team’s offering. It uses automated speech recognition (ASR) and natural language processing (NLP) to decipher audio.
This is then converted to text within seconds, with Trint saying that a first draft, time-coded transcript can be 99% accurate.
Video or audio files, including mp3, wav and mov, can be uploaded straight into the software, while live transcription is also available.
The transcribed files can be easily searched, while a variety of highlight colours can be used to pinpoint specific quotes. Transcripts and ideas can also be easily shared between team members.
Saving time at Handelsblatt Media Group
One beneficiary of Trint’s focus on saving time has been Handelsblatt Media Group (HMG) in Germany.
Florian Hueckelheim, head of business editorial at Solutions, part of HMG, said in case-study on the Trint website: “If I compare Trint to the way I used to do interviews, I’m four times faster in bringing content from a recording or my notes to an article that can be sent to colleagues for review.”
Hueckelheim and his team now use Trint whenever they need to turn audio into text. This means a 40-minute interview between two or three people can take a fraction of the time.
“That way it takes me 30 minutes instead of maybe 120, 150 minutes,” he said. “It’s nice to see that there’s a solution that can handle this for us.”
The group’s editors also use Trint to transcribe their audio recordings into German, which is one of the 30-plus languages that the software can accommodate.
According to Hueckelheim, the German transcripts are “good enough to be very useful”, even if people drop a syllable or make up words during interviews.
“If you read a line that looks strange, you can easily jump to that sentence and listen to the recording again to correct it,” he explained.
The translation function has also benefitted the Center for Investigative Reporting (CIR), which enjoys an enviable reputation for journalism and ground-breaking storytelling.
Translating into Spanish and Russian have been among the requirements at Reveal News, the CIR’s national radio show and podcast.
Quoted on the Trint website, CIR senior editor David Ritsher said: “With Trint, we are free to transcribe and translate much more material, which helps us easily find better excerpts to use.”
He also praised the Trint’s ability to integrate with other platforms: “One of the critical strengths of Trint is its really powerful API,” he explained. “We were able to integrate Trint with the Media Asset Management system we use to overcome some internal challenges.”
Being able to collaborate effectively and reducing the time wasted are seen as additional benefits of using the platform. At CIR Trint as enabled reports to work more efficiently and focus on the content of reporting rather than on menial tasks associated with processing raw recordings.
San Francisco Chronicle: Security benefits for investigations
At the San Francisco Chronicle Trint is widely used by the podcast team: Every interview is put into Trint as a first step.
It helps the team edit their interviews more precisely by reviewing the transcript to decide which sections to cut and which to highlight, and how to fine-tune each episode.
It is also valued as security tool used by the title’s investigative team to ensure that the accounts of sources are kept securely
Feldberg also values its role as a security tool that has helped support the work of the organisation’s investigative team over the past year.
Trint acts as a single secure location for interviews with sources. It means that on sensitive stories, interviewees can be assured their data is safe.
Trint is fully certified to ISO 27001:2013 as set by the International Standards Organization to provide a global standard for information security management systems (ISMS). This is considered the platinum standard for data security as verified by SAM (US government vendor), the UK Crown Commercial Service (CCS) and Cyber Essentials.
Trint mobile app enables lives transcription in more than 30 languages
Trint has just launched their new industry-leading collaboration tool Trint Mobile Live. Mobile Live allows anyone working in a fast-paced news media or business environment to capture and transcribe audio from their phone and share with colleagues in real-time. All users need is the Trint mobile app installed for iOS or Android to transcribe it live in more than 30 languages. Mobile Live is an exclusive feature for Trint enterprise plans.
Free trial period for publishers
Trint is so confident that journalists and publishers will love the platform that it’s offering a week’s free trial without any need to share credit or debit card details.
There are starter and advanced packages available. Depending on the option chosen, users can transcribe, collaborate, create, share and publish their work.
In addition, there’s an enterprise option for organisations that have more complicated security, management and workflow requirements.
The post How PBS, San Francisco Chronicle and others are using AI-driven speech-to-text software appeared first on Press Gazette.