Case Study: Automating Tagging for Enhanced Translation and Voiceover Accuracy

Explore how Cloud Combinator used a Large Language Model and wider AWS services within AWS’s Bedrock to automate the tagging process for voxANN, improving efficiency and accuracy in their translation and voiceover platform.

About The Customer

voxANN is an innovative company dedicated to bridging the gap between written text and spoken word through a combination of human expertise and AI technology. At the time of writing, they are close to launching a cutting-edge tool for the Language Services industry, helping translation agencies and multilingual voiceover studios streamline their production processes. 

voxANN's platform offers precise control over the different elements involved in translation and voiceover projects across multiple languages. It allows language professionals to synchronise and amend scripts, generate performance directives, and ensure culturally relevant translations and delivery. Their innovative approach aims to save time and achieve unparalleled accuracy and synchronicity between languages.

Customer Challenge

voxANN’s proposed project focused on improving efficiency - and the end-user benefits that would come if they could streamline their translation and voiceover process. At the core of their challenge, there was a human process - the manual tagging of words, phrases, and subsections in transcripts. This process was not only time-consuming but - being human-driven - also lacked a systematic approach, leading to inefficiencies and possible inconsistencies.

The primary challenge was the random generation of tags without any structured system. This manual method required individuals to go through transcripts and apply tags, which was both labour-intensive and prone to errors. If this issue were not addressed, it could lead to delays in project timelines, increased operational costs, and potential dissatisfaction among their clients, which include translation agencies and multilingual voiceover studios.

What’s more, the inability to automate the tagging process hindered voxANN's ability to fully leverage their AI technology, limiting the potential for scalability and efficiency improvements. Addressing this challenge was crucial for maintaining their competitive edge in the Language Services industry and ensuring the delivery of very high-quality, truly synchronised translations.

Partner Solution

To address voxANN's challenge of manual and inconsistent tagging, The Cloud Combinator team created a solution primarily using AWS Comprehend alongside a small selection of other AWS services. The primary goal was to automate the tagging process, improving accuracy and efficiency.

Key Components of the Solution:

AWS Comprehend Custom Classification Model: Working with voxANN, the Cloud Combinator team created a dataset suitable for training a custom classification model in AWS Comprehend. This model was designed to automatically apply tags to words, phrases, and subsections based on the provided dataset. The trained model processes phrases sent through an API, returning appropriate tags, thereby eliminating the need for manual tagging.

• AWS Lambda: Serverless computing was employed to handle the data processing workflow. Lambda functions managed the flow of information, ensuring efficient handling and processing without the need for dedicated server management.
• Entity Detection: The Cloud Combinator team integrated entity detection within the pipeline. This detection capability identifies and tags entities such as names, companies, amounts, dates, numbers, addresses, and emails, enhancing the overall tagging accuracy.
• Implementation and Integration: As per the final steps of our Cloud Combinator program, comprehensive training was provided to the voxANN team, covering operational aspects and troubleshooting to ensure seamless adoption of the new system.

Results and Benefits

The implementation of Cloud Combinator's solution using AWS services successfully addressed voxANN's challenge of manual tagging and human-error inconsistencies. By implementing the custom classification model in AWS Comprehend, voxANN has significantly improved the efficiency and accuracy of their tagging process for their customers.

The solution automates the previously manual tagging process - therefore saving considerable time and reducing the workload for voxANN’s customers. This automation has led to accurate and consistent tagging - enhancing the quality of translations and voiceovers. The ability to automate this core part of the voxANN services not only streamlines operations but also ensures that the tags applied are precise and reliable.

It’s not just future efficiency that the Cloud Combinator project unlocked though - the voxANN team saved 75 hours of developer up-skill and development time - while fast-forwarding their Amazon Comprehend knowledge and allowing the team to concentrate on building their prototype solution, supported by the Cloud Combinator team’s work.

voxANN is now also able to handle more projects simultaneously with faster turnaround times for customers. By eliminating manual errors, the reliability of the tagging process has increased, ensuring high standards of output. This improvement in efficiency enables voxANN's customers to meet client deadlines more effectively and maintain a high level of service quality.

The scalable nature of AWS services also means that voxANN can easily handle an increasing volume of data and projects without compromising on performance. The integration of entity detection refines the tagging process even further, making it more comprehensive and effective. As a result, voxANN is well-equipped to manage growing demands and expand their service offerings - enabling their customers to remain competitive in the Language Services industry.

Why Cloud Combinator?

Cloud Combinator’s Advanced AWS Partner status and bleeding-edge working knowledge of AWS Comprehend made the team a stand-out choice for voxANN’s AI tagging project. While creating an on-brief solution that met technical requirements was obviously essential, Cloud Combinator’s broader business knowledge also meant understanding how the process would enhance voxANN’s offering - making it possible for the team to recommend ways of implementing the solution that would allow for scalability while remaining budget-friendly into the future.

Cloud Combinator’s proven AI & ML Project Accelerator process also helped to ensure voxANN’s requirements were met. As well as plotting a swift and clearly-defined timeline for the project, AI & ML Project Accelerator meant the team could effectively leverage AWS funding for voxANN, helping to present a proof-of-concept to the business’s decision makers without incurring additional costs.

Client Satisfaction

“Through a series of discovery meetings, Cloud Combinator took time to understand our business and the solutions we were aiming to build. From this, they were able to identify where their skills could augment our own and formed a solution around one of the AI enabled features we’d planned to implement. The teams at Cloud Combinator proposed a solution using Amazon Comprehend and a few weeks later we had a demo, proving we were on the right track. Happy with the progress, Cloud Combinator then built out the solution and final delivery was swift and packaged up in a way that dovetailed nicely with our early stage prototype. Integration was a breeze and propelled us forward to into the world of AWS AI services!”

Robin Tong, voxANN