The Role of Humans in the Language Industry While Machine Learning Evolves

Chetna Aggarwal (MATLM 2020)
Chetna Aggarwal (MATLM 2020)

How Artificial Intelligence Changes Your Decision-Making Process

The success of the Georgetown-IBM collaboration in the 1950s lead researchers to think that machine translation will replace human translation in only a few years (Kelly, 2014). Among the first things that I learned after I started immersing myself into the translation and localization industry was that this statement is not true and that machine translation will not pose a threat to the future generations of translators (Nimdzi, n.d.). A major reason for this is the fact that this technology and one of its subsets,  artificial intelligence (AI), are still developing and therefore a fully automated, high quality and unrestricted translation is currently not feasible without human intervention. How can businesses and other stakeholders in the language industry then prepare themselves for this evolution to maximize their returns and simultaneously offer the best possible service within a short period of time?

Machine Translation and the Business Model Canvas

I would like to use the business model canvas developed by Alexander Osterwalder (see appendix) to illustrate my thoughts on these processes. The reason I opted for this strategic model is because it organizes the decision-making process of a project manager (PM) in a logical and organized way and depicts all relevant areas of a business.

First, the decision-maker, who may be a freelance translator or a PM in a company, needs to consider his or her value proposition. In other words, how will machine translation change the company output that will be used by the end-consumer? There is tremendous potential and rapid growth that PMs need to be aware of, which can overall be described as a continuous exposure to AI in the next years to come (O’Dowd, 2019). What will change for the client? While companies become more agile by using new technologies, customers benefit from a more interactive, customized and high-quality product or service (Nimdzi, n.d.).

A second area that illustrates the supply side of the business model canvas includes the key partners (who am I working with), the key activities (what do I need to do) and the key resources (what do I need) that define how a company intends to offer the value proposition. While AI offers many new opportunities to grow, it also requires PMs to be on the constant lookout for new technologies and train their work force accordingly to finally create competitive advantage. The introduction of neural machine translation in the late 2000s has already been a major breakthrough in the industry that will allow PMs to allocate company resources more efficiently. Time spent on simple or cumbersome translation processes can instead be used in areas where problem-solving skills, creativity or innovation is needed (Nimdzi, n.d.).

To become more efficient, companies need a reliable labor force and trustworthy language service providers as partners who produce or contribute to the desired output. PMs need to make decisions on which tools and features they require and what kind of staff is used for each and every step in the translation and localization process. These changes, if implemented correctly, can not only reduce production time and costs but also increase the translation output. However, with the rise of modern technology, data security has become an even bigger issue. In other words, using machine translation can also harm a client when confidential data in a translation memory is disclosed or deleted. Hence, a part of all key activities and resources are the methods and tools (i.e. blockchain) used to protect sensitive data.

A third area that the business model canvas is concerned with is the demand side and therefore, the end-consumer. How do I maintain the relationship to my customers (customer relationship), who is my customer (customer segments) and finally, how do I sell my product or service (channels)? While machine learning does not directly affect these areas, it is still very important to consider these factors because bad machine translation technologies are the reason why companies choose to consult professionals to complete a job. Customers are the ones who are directly affected and the reason why translations are completed in the first place. Using modern technologies to complete a task and to store data with translation memories or term bases is ultimately beneficial to the end-consumer. Post-edited translations and / or layouts are saved and can be reused for future jobs, for which less time and money will have to be invested.

Finally, the business model canvas discusses the cost structure (how much do I need to invest) and the revenue streams (how much do I need to earn to break-even). The combination of the two result in a minimum viable product that we are able to bring onto the market. Machine translation, especially AI technology and neural machine translation are better technologies than any of their predecessors. However, the initial investment is also much higher, considering the fact that it is new technology still in the development stage (Nimdzi, n.d.). Currently, such an investment makes only sense if the amount of translations to be completed is big enough to sustain such an expense (i.e. user manuals as opposed to websites). However, is the company operating in multiple countries and translating documentation into numerous languages, it might make sense to invest in AI and the required training to operate these machines.

All of these points and the totality of the business model canvas show which decisions and processes a PM has to go through when considering machine translation in his or her business. It starts with defining how AI will enhance the end product, followed by identifying which part of the business is concerned and choosing whether these new technologies can in fact maximize the company turnover. In the case of a positive inclination for AI, a company scan needs to be completed to conclude whether a minimum viable product can be produced after having addressed all issues around the value proposition that are essentially the various areas of the business model canvas.

What Does This Mean for the Future of the Industry?

The production of the first machine translation in the 1950s was merely a start to what might become fully automated translation. It is evident that computer-assisted technology is required in a globalized world, where a vast amount of data and information crosses borders within milliseconds and the need for translated text has become more important than ever to reach the largest possible audience. However, even though technology is ubiquitous, it is clear that human translators are indispensable. Though the role of the translator is slowly shifting towards one of an editor (especially with neural machine translation), the complexity of a language cannot be mastered by a machine (Kelly, 2014). The primary reason for this is the context of a document and therefore, the translation quality. English words such as “get” or “run” have a myriad of meanings that cannot all be understood by machines, which is mostly the case with idiomatic sentences or similar texts for different industries. Therefore, the translation and localization landscape continues to become more technical, however, it won’t exist without human involvement. Translation and technical priorities will finally, allow for a different resource allocation that will increase the final output.


Kelly, N. (01/09/2014). Why Machines Alone Cannot Solve the World’s Translation Problem. Huffington Post. Retrieved from

Nimdzi. (n.d.). AI Meets Localization. Nimdzi. Retrieved from

O’Dowd, T. (01/01/2019). Localization tech predictions for 2019. MultiLingual. Retrieved from

Osterwalder, A. (2019). The Business Model Canvas. Strategyzer. Retrieved from


Leave a Reply