Automation

Machine language translation for automation is disrupting global service delivery

Share this post:

Key Points:

  • Restrictive offshoring policies are putting unprecedented pressure today on companies to reinvent their business models and look for alternatives to operate their back-office processes.
  • Automation using customized machine learning language translators are slowly emerging as alternatives.
  • Cognitive language translators—like speech-to-text or text-to-text capabilities—are expected to create a renaissance in the global services delivery model.

Whether you call it cognitive computing or artificial intelligence (AI), it has evolved to become an enterprise priority. Having realized its vast potential, enterprises have started using cognitive computing not only as technology to derive insights and intelligence from enormous amounts of unstructured data, but also as a core automation enabler in more tasks previously performed by humans. Implementing cognitive computing-based automation (also called cognitive automation) at the core of your operations helps digitize, integrate, and reduce friction in front- to back-office business processes end to end. Its applicability goes beyond horizontal processes such as finance, procurement and human resources to industry-specific back-office processes. A recent IBM Institute for Business Value (IBV) study (Accelerating enterprise reinvention: How to build a cognitive organization) claims that 50% of global chief executive officers (CEOs) plan to implement cognitive computing by 2019.

Of the many cognitive automation applications that are being deployed, language translation has gained momentum and become a critical application for several reasons. Machine language translation technologies such as such IBM Watson® Language Translator is poised for rapid global growth in the next 2-to-3 years. We are already seeing AI translation technologies capture the headlines in the news. For example, a recent press release stated that Australian start-up taps IBM Watson to launch language translation earpiece. The embedded AI in the earpiece supports translation in eight languages in five seconds. Last month, Facebook announced that it is using AI to make language translation nine times faster than current tools.

It’s no secret that companies are eager to launch AI-powered machine language translation services. According to a research from Oxford, AI will surpass human capability by 2045 across several domains, and language translation by 2024.

Language translation challenges in global service delivery

Restrictive offshoring policies are putting unprecedented pressure today on companies to reinvent their business models and look for alternatives to operate their back-office processes.

Language barriers across geographical boundaries and the rising costs of skilled hires at nearshore locations supporting businesses for language needs also put pressure on companies to seek alternate approaches. Enterprises—whether operating through in-house or outsourced environments—are increasingly challenged with:
• Using nearshore centers to deliver economies of scale as opposed to offshore delivery centers
• Hiring language-specific resources at offshore centers, which doesn’t offer a scalable solution
• Using “cheat sheets” to meet translation needs, which limits the communication and impacts operational effectiveness

Affordable automation through machine language translators

Language translators translate text from one language to another, and AI-powered translators offer a distinct capability. Natural language processing and translation using a cognitive system can now produce conversational translations with a fair degree of accuracy.

For example, IBM Watson Language Translator service uses massive amounts of data and increased processing power to deliver more accurate translations. The service offers multiple domain-specific models you can customize based on your unique terminology and language.

How machine language translation be deployed within finance processes

This application automates language translation tasks in invoice processing in finance operations to help improve practitioners’ productivity at global delivery centers (such as India or Philippines)¹. It also helps reduce the dependency on regional teams and improve the efficiency of the offshore teams. The application provides security-rich and domain-specific language translation service between multiple languages in scope. The translation service can also be integrated into the overall enterprise resource planning (ERP) workflow to facilitate process automation. There are two components that work hand in hand to optimize the process specific translation service:
• Translator application for agents translates inflow comments from stakeholders and sends back queries in languages the stakeholder can understand.
• Trainer application reviews quality of translation and provides input, which helps to both customize the underlying model to maintain good performance and ensure domain relevance.

Conclusion

Machine language translation services can provide deep cognitive capabilities. But the current translation applications typically have the capability to offer phrase-by-phrase translations, and are prone to error. These systems don’t cater to the nuances and subtleties of the language.

Automation using customized cognitive language translators are slowly emerging as alternatives. In the near term, cognitive language translators—like speech-to-text or text-to-text capabilities—are expected to create a renaissance in the global services delivery model. Of course, new skills will be required by humans to operate and collaborate comfortably with these next-generation applications.

References
1 “Cognitive Computing in Action to Enhance Invoice Processing with Customized Language Translation,” Ying Li, M. Muthiah, A. Routh and C. Dorai, IEEE International Conference on Cognitive Computing (ICCC), 2017.

 

Learn more about how automation can help your organization transform business processes.

 

 

Add Comment
No Comments

Leave a Reply

Your email address will not be published.Required fields are marked *