Artificial Intelligence and Localization

Artificial intelligence is more than a matter of technological innovation for localization

Artificial intelligence is a game changer for the content industry in general and the language industry in particular, in terms of both technology and services. Unsurprisingly, it has had a vast and continued impact on localization people, processes, and technology. In the early days of artificial intelligence (AI), machine learning (ML), and machine translation (MT) offered a basic and automatic conversion of text, such as glorified versions of online dictionaries and glossaries. That time is over. ML and MT are now propelled by neural networks that bring them closer to functioning and behaving like human brains. Although AI is getting closer to human behavior, it cannot mimic it fully, and therefore, it has not yet replaced human beings. As far as localization is concerned, ML and MT have made significant progress over the past few years. These two AI branches have drastically changed how localization teams and individuals operate. They have also influenced how customer experiences are delivered effectively in local markets. Creating value and shaping local experiences are the purposes of AI-driven localization, as evidenced by the rise of multilingual conversational AI. Chatbots and voice assistants make machine-to-human (M2H) interactions more common and natural. AI-driven localization makes them more relevant and engaging with content in the customer’s native language.

AI turns localization into a profit driver 

Today, ML and MT make it easier to create, capture and measure localization value for the benefit of global businesses and local customers. AI contributes to making localization a tangible business enabler and a recognized profit driver by accelerating intelligent automation and human elevation – which are urgently needed to be successful. It is even more important at a time when digital globalization is re-calibrated to consider fast-changing client requirements and to make faster decisions about how to meet them. As estimated by the Brooking Institute, there will be more than 4.7 billion middle-class and affluent consumers in non-English-speaking regions by the year 2030. Their language will be at the center of their experience.

AI-driven localization is boosted by research and investment internationally

ML and MT are branches of AI into which public and private organizations invest the most time and money globally. The Technology Innovation Institute (TII) of Abu Dhabi is an example of this trend. It is a global research center dedicated to pushing the frontiers of knowledge in a variety of advanced scientific fields, including natural language processing, which is a vital component of the AI sector. TII recently unveiled the world’s largest Arabic language processing model, Noor, which is designed to boost the development of tools like chatbots, language translation, and sentiment analysis across social media.

AI makes a difference in localization resources

AI enables localization people to work faster and more effectively regardless of their role and responsibility in the localization supply chain. These resources can avoid several repetitive and low-level tasks, therefore freeing up more time for value-creating and value-adding activities, enhancing localized content, and delivering multilingual content faster than ever before. ML and MT elevate the role of localization people by including them in the customer value chain and amplifying their contribution to the linguistic, cultural, and functional effectiveness of localized content.

Post-editing content means improving and enriching AI-driven localization

Machine learning and machine translation can ensure that the required levels of accuracy and consistency are delivered in the initially produced output so that linguists and project managers can edit it accordingly. They can then focus on enriching localized content to increase its level of fluency, fluidity, and customization. Some light post-edits may be required for content that does not have to be published, whereas heavy post-editing matters for client-facing materials or personalized content. Localization engineers and testers also benefit from AI-driven localization. They can automate procedures and protocols that do not require much human attention and intervention, especially regarding highly structured and repetitive content. Segmentation, testing, and validation can then be accelerated and controlled by AI, while engineers and testers can perform value-adding tasks for sensitive and unstructured content. Instead of working with files, they can manage content as segments, empowering them to collaborate with content creators. ML and MT effectiveness eventually depends on the quality of language data used to train them. 

Putting and keeping humans in the loop with AI-driven localization

AI-driven localization is as efficient as the combination of human and machine intelligence makes it. Striking this balance is one of the major challenges since it defines the best leverage of ML and MT for localization and how capacity and capabilities must be adjusted accordingly. Bear in mind that translation and localization are trickier than they may seem. Linguists, project managers, and localization engineers must take up linguistic, cultural, legal, functional, and ethical challenges that require human judgment and action. Localization takes more than converting words into one or more foreign languages. In the digital age, content includes text, pictures, voices, and sounds that speak to the mind and hearts of local customers. Therefore, localization resources are mobilized to team up with the machine to ensure that the final localized content is consumed naturally and engagingly. 

Overlooking or underestimating the human-in-the-loop requirement can break content effectiveness and customer experiences, even more in regulated industries where human life, security, and wealth are at stake. 

AI-driven localization makes a difference in localization processes

ML and MT have taken automation to the next level by making it intelligent. Specifically, it means that AI has made automation more scalable, capable, and adaptable to handle increasing amounts of content, a vast array of content types, and fast-changing delivery expectations. It has enabled localization to deliver more value with less waste of time and money. 

Automating workflows in AI-driven localization 

AI works faster than humans. ML and MT process incredible amounts of content at unparalleled speed and help automate – and therefore accelerate – localization workflows. One of the reasons why AI changes the face of localization is because it can be deployed throughout localization workflows –from translation to quality assurance. In traditional localization, linguists and project managers go through a number of steps and iterations, including translation, localization, review(s), and final delivery. Linguists must localize content file by file, and project managers must manage their tasks in the same fragmented and sometimes unsynchronized way. With the AI-driven approach, localization resources work with language data in the form of content segments, allowing multiple resources to work on the same content simultaneously. Localized content can be captured and leveraged faster to train ML and MT engines, and it can be delivered more safely when and where it matters for local customers. Workflows can be significantly streamlined and more aligned with customer journeys. This way, ML and MT turn localization supply chains into experience value chains.

Managing AI-driven localization with a well-designed and validated data strategy

AI learns faster than humans. ML and MT require a robust data management strategy as they need large amounts of data to be trained in a timely and accurate way. Algorithms must be trained accurately and holistically to be able to work with any type of content. Data is needed to inform the machine’s intelligence to cover all use cases that may come up during localization. It enables this intelligence to work correctly to meet requirements accurately. The data sources must be carefully selected, and data must be organized so that the algorithm can identify it to meet requirements as expected. Finally, the ML and MT training process can start and be evaluated in a staging environment before going live when it proves effective. Training AI is an ongoing process, so it should never stop to avoid decreasing the level of machine-translated content quality. The output quality is measured according to performance indicators that capture the edit distance between the machine and human translation. The fewer edits needed, the more accurate the machine translation. BLEU scores are considered in most MT quality assessments and may vary according to language combinations, especially between European languages and Asian languages

AI-driven localization makes a difference in localization technology

Technology enables people to use processes timely and cost-effectively. AI has triggered a change in language technology as localization people have had to deal with evolving processes. ML and MT have disrupted how localization technology is designed and developed by solutions providers, as well as how it is implemented and embedded by solutions buyers. 

Creating a localization ecosystem optimized with AI

As AI accelerates the overall localization lifecycles, it goes beyond the usual combination between a translation management system (TMS) and a computer-assisted translation (CAT) tool, which prevails in traditional localization. ML and MT are the foundation of an AI-driven localization ecosystem that covers the end-to-end content supply chain. AI plays another critical role, i.e., creating localization-ready source content. AI-driven localization ecosystems typically include components dedicated to content management, translation management, vendor management, digital assets management, etc. AI ensures that these components are well connected or integrated and process information quickly and safely. Comprehensive localization ecosystems pave the way to new paradigms in localization operations that extend the localization scope and involvement to cross-functional and multidisciplinary teams.

Embedding AI-driven localization technology in corporate environments

The continuing ML and MT progress in reliability and agility is increasing the accessibility to a broader audience and becoming more applicable to wider domains. While the heavy users of these technologies remain localization people, the use of customized ML and MT services via programming interfaces (APIs) has gradually spread for the past few years. Connection and integration options with infrastructures allow the use of MT for various purposes such as – yet not limited to – email, social media, chat features, or knowledge bases. The expected machine translation leveraged value should be balanced with cost implications and technical challenges specific to each organization.

It is essential always to consider the beneficial use of AI in localization with two objectives in mind. Firstly, ML and MT should create value through automation, acceleration, communication, and integration within organizations. Secondly, ML and MT must be leveraged, integrated, trained, and customized to improve organizations’ delivery of local customer experiences.