IABM Journal

Representing Broadcast & Media Technology Suppliers Worldwide
Articles covering a range of key topics and themes
In depth analysis of the latest trends

Automated Transcription and Localization With a Human Touch Can Transform Your Media Operation

Russell Vijayan, Head of AI Products and Services at Digital Nirvana

Mon 18, 04 2022

Enterprises and media companies rely on transcripts to create a written record of movies, TV shows, meetings, speeches, phone calls, newscasts, training videos, and more, thereby making the content searchable, discoverable, and accessible for internal and external uses. And they need only translate the transcripts to reach people in different parts of the world who speak different languages. Content localization has become more important since the start of the pandemic as more and more companies reach across continents.

 

But for the vast majority of these companies, transcription is not their core business. Transcribing the content and translating it to multiple languages in-house is an administrative hassle that takes time and energy away from what they really need to accomplish.

 

Either they have to implement the technology and find, hire, and manage a team of qualified people to handle the process themselves, or they must allot staff to find, hire, and manage transcription/translation providers. That requires multiple personnel to assign jobs to vendors based on capacity and to monitor incoming transcription files for on-time arrival. It also means running a full quality-check operation to make sure transcriptions and translations meet standards for accuracy, style, and compliance, especially with specialized topics.

 

Sometimes transcripts must be integrated with the audio or video in several output formats. In that case, there is an engineering requirement. Either they have to hire a development company to do this, or they have to use internal engineering resources to build something that can create a transcript and then convert it into different formats.

 

Some companies, such as news broadcasters, have multiple, continuous feeds coming in a time that must be transcribed and translated in real time. Others, such as movie studios, require transcripts for stored content. Still others, such as financial data providers, have peaks and valleys in their transcription/translation needs. Accounting for that demand is another challenge, especially when outsourcing. It’s hard to keep freelancers or small-scale vendors busy during slow times, and many vendors don’t have the capacity to ramp up quickly during an onslaught.

 

Enter AI-driven transcription and translation services paired with human experts.

 

A solution like this relies on dedicated language and content specialists. This team, aided by cloud-based technology, generates highly accurate transcripts, captions, and text translations within tight turnaround times.

 

It works like this:

 

Companies initiate the process by transferring a video or audio file through multiple data transfer options including a portal, API integration to existing workflows, MAM, PAM, or even a file sharing platform with a request to transcribe and localize. For example, let’s say the source audio is in English, and the request is to provide transcripts in English, Latin Spanish, and French.

 

From there the automated system will ingest the file and use speech to text and natural language processing to generate an English transcript. After the speech-to-text generation, there are two layers of manual quality checks to make sure everything is 100% accurate. They system has pre-set rules and NLP algorithms which are used to convert them to closed captions. Once that is done, the entire transcript goes through an automatic machine translation process, and then it goes through another two layers of QC by humans — one by a language expert and another by an in-country language specialist in Latin Spanish. The same process happens simultaneously in French.

 

Before creating the final output, the system performs an automatic process to check for crucial elements, such as style conformance and missing information. If there is any issue, the system will alert and make it impossible to upload transcripts back to the client’s portal until a supervisor has reviewed and approved the output. The system automatically generates the output in all required formats and delivers it back to the company’s portal.

 

Post-delivery auditing — whereby an auditing team evaluates a percentage of all delivered files based on various parameters — helps to ensure the system and quality continuously improve.

 

This solution delivers:

  • Consistent quality — Multiple levels of automated and manual QC ensure all boxes are ticked before the final output reaches the customer.
  • Reliability — A global presence, coupled with technological innovations, keeps the operation running 24/7 to meet uninterrupted service deliveries.
  • Ability to ramp up — The entire operation is set up to handle peaks and valleys in terms of the amount of data to be delivered.
  • Localization with pinpoint accuracy — In-country translators catch slang terms, nuances, and the essence of the message that speech to text and even other humans could miss.

 

Gone are the days of building in-house transcription capabilities from the ground up or managing multiple vendors. Instead, many companies are using turnkey transcription and translation services built for scalability to handle the real-time transcription function from beginning to end.

Translator professional and languages education concepts. Hand reach for notebook and symbol of translation (speech bubble with arrows and abstract text) and globally important languages.

Case in Point

One example is a global company that aggregates the expertise of professionals around the world in a variety of industries. Private equity, investment funds, management consultancies, corporations, and nonprofits rely on this information to help make better decisions.

 

At any given time, there could be several industry-specific discussions happening with different experts from different viewpoints and in different geographies across the globe — which require transcription and localization before the company can put them to use. The company’s algorithms derive insights from those transcripts. Also, in the end, the company includes the transcript in desired languages in the package that goes to its subscribers so that they can take elements from it for things like research reports or to mine additional insights.

 

This company tried starting an in-house team on its own but got bogged down by the administrative complexity of hiring, management, and quality control. It required a large team to oversee multiple freelancers and small-scale vendors — a team that spent a lot of time following up with freelancers and vendors that hadn’t delivered on time. It was tough to find vendors that could handle spikes in volume as well as the capability to localize into different languages. Even worse, sometimes freelancers or vendors would opt out mid-project because of difficult audio or complexities involved in localizing. Because of these delays, the company was missing time commitments to its customers.

 

Results

After implementing the turnkey solution, this global insight provider’s transcription operation was transformed.

  • It greatly reduced the number of administrators and time required to deal with transcription.
  • It eliminated the need for multiple vendors.
  • It helped transcribing in 30 different languages and localize all to one common language for insight generation.
  • It scales up and down to handle fluctuations in volume with no loss of quality.
  • It consistently delivers accurate transcripts well within the company’s turnaround times, even during peak periods, so the company can meet its customer commitments.
  • It delivers accuracy the company can trust without performing its own quality checks, saving time and effort.

Search For More Content



X