December 2016
By Arle Lommel

Image: © Le Moal Olivier/123rf.com

Arle Lommel is a senior analyst with independent market research firm Common Sense Advisory (CSA Research). He is a recognized expert in quality processes and interoperability standards. Arle’s research focuses on technology, quality assessment, and interoperability.

arle[at]commonsenseadvisory.com
www.commonsenseadvisory.com
Twitter: @CSA_Research

Language technology leads to growth for all

Machine translation enables organizations to dramatically increase the amount of content they translate and the languages they support. And, as the technology improves and matures, it has gained acceptance among content-producing organizations and language service providers.

Machine translation (MT) is a permanent – and highly controversial – topic for organizations that require translation as well as for the language service providers (LSPs) who work with them. Common Sense Advisory’s (CSA Research) examination of MT showed that it is poised on the edge of a tremendous growth curve that will take it from a niche solution suitable primarily for large enterprises to a mainstream one for companies of any size. This article examines the changing role of MT and how this affects language service providers (LSPs), enterprise content developers, individual technical communicators, and human translators.

Content growth renders MT inevitable

CSA Research estimates that the global capacity of all professional translators today could address only 0.00000000009 percent – less than one-billionth – of the content generated every day. Even though much of that content consists of non-translatable data – such as sensor readings and machine-to-machine communication – far less than one percent of the content will be translated that, in an ideal world, would appear in multiple languages. Add to that the fact that most translations are made into a small number of economically important languages that leave a big part of the world on the outside, as strangers looking in on a wealth of materials they cannot access – a phenomenon known as language blocking.

If organizations are to close the gap and provide a positive customer experience to individuals who purchase their goods and services – no matter where they live – they will need to turn to technology. Human translators cannot possibly close this ever-growing gap – even if a massive influx of new translators enters the field. Fifteen years ago, globally operating companies addressed at most a few dozen languages, but today they support 120 languages or more, including some for which very little professional translation capacity exists. Given the trend for these numbers to steadily increase, it is apparent that LSPs will need to embrace technology to meet the needs of their clients.

How will LSPs meet the challenge?

LSPs are aware of these trends, and also respond to increasing client requests for MT-related services. Over time, their attitudes have shifted from opposing the new technology to cautiously experimenting with it and to finally accepting it. Today, many embrace it.

CSA Research asked CEOs of leading LSPs how they view and use MT in their companies. The vast majority of them – 92 percent – stated that it played a role. Just eight percent intended to stick with human translation only. Although the survey focused on larger LSPs, thereby overstating the level of adoption compared to smaller LSPs, by 2019 the majority of both enterprises and LSPs will use MT for at least some of their international content production needs. In fact, the largest growth market for MT production is currently among small and medium LSPs.

Image 1: Only nine percent of businesses intend to stick with human translation only.
Source: CSA Research

 

LSPs generally prefer to offer post-edited machine translation (PEMT). This refers to translations where professional linguists review MT output and correct it to meet defined quality targets. They do so by correcting grammatical and translation errors, either with the goal of making the text accurate and understandable – so-called "light" post-editing – or of making it indistinguishable from human translation – "heavy" post-editing.

The current growth of PEMT – coupled with surging demand for machine translation services in general – is generating a massive shift in how LSPs produce translation and deal with technology.

LSPs embrace MT to drive growth

Based on the importance the respective CEOs assigned to MT, CSA Research grouped LSPs into two roughly equal groups: aggressive adopters, who see MT as a "crucial" or "very important" part of their business plans (44 percent), and more conservative users who treat it as "somewhat important" (47 percent). Many of the latter group offer PEMT only in direct response to client requests, but do not otherwise market or promote it. Not surprisingly, companies that appeared in CSA Research’s annual ranking of the 100 largest LSPs were more likely to adopt an aggressive approach, as were many smaller up-and-coming companies.

The analysis correlated these groups with their companies’ compound annual growth rates (CAGR) from 2013 through 2015 based on reported figures in CSA Research’s annual Global Market Survey. Overall, respondents grew at a healthy average of 5.5 percent per annum, but these growth rates exhibit a particularly strong and statistically significant correlation with MT strategy. LSPs with an aggressive implementation approach grew 3.5 times as much as their more cautious competitors – 7.9 percent CAGR versus 2.3 percent.

Image: © LSPs with an aggressive MT implementation approach grew more than their conservative competitors.
Source: CSA Research


Is this difference just a result of the fact that the largest LSPs are more likely to take an aggressive approach to MT and, independent of that, they also happen to grow faster than their competitors? No. In fact, CSA Research found the opposite effect. Machine translation strategy is a much better predictor of growth rates than company size. Although the largest providers do grow faster than smaller ones (6.6 versus 4.4 percent), this difference is not statistically significant. To the extent that the larger firms grow more quickly, CSA Research attributes some of that increase to the existence of more aggressive MT implementers in that group.

Although these results do not indicate that MT adoption leads directly to faster growth, they do show two circumstantial connections:

  1. LSPs experiencing rapid growth are more likely to use machine translation, and
  2. Those with a technology-centric approach are more successful in managing rapid revenue increases.

CSA Research also ascribes some of the cause for the rapid growth of aggressive implementers to their early-adopter advantage. They can provide more cost-effective services for which there is pent-up demand. This difference is likely to fade as increasing numbers of LSPs move to work with MT and it becomes a standard part of provider toolsets, but for now, MT represents an opportunity for those who seek to grow.

Enterprises can leverage MT for global growth

The advantages for LSPs that are willing to work with MT are clear, but the ultimate beneficiaries are enterprises that produce content. Examination of the market reveals that – after a period of market uncertainty – prices for heavy PEMT are slowly stabilizing at around 65 percent of the cost of full human translation. This reduction in cost makes translation more attractive and enables companies to address more content or add additional languages.

In the next three years, all types of translation – including "pure" human ones – will see substantial volume increases. Compared to present baselines, the fastest growing segment will be PEMT (105 percent increase in volume from 2016 to 2019). Unedited ("raw") MT will lead in terms of sheer volume, but much of this will be for comparatively low-value text that would otherwise remain untranslated. For important texts, PEMT becomes increasingly an option. CSA Research’s examination of what enterprises are translating shows product documentation, online Help, FAQs, and knowledge bases as attractive targets for PEMT. However, marketing – which requires a deep understanding of the culture and characteristics of its audience – will remain in the hands of human translators for the foreseeable future.

Raw MT, by contrast, provides opportunities for companies to deal with customer engagement content – blogs and comments, user-to-user forums, customer reviews, messaging, group discussions, and social media posts – that they would generally not translate. These types of content have uncertain value, a shelf-life that is often too short for slow human translation, and little funding for translation. For scenarios where they can live with results that may fall short of what human translators can deliver, raw MT provides a way for organizations to expand their coverage and engage customers more effectively.

Technical writers influence MT success

These market changes will also affect writers. MT requires communicators to change their habits. Statistical machine translation, which dominates the field today, does not "understand" the text it translates. Instead, it relies on detecting similarities between new content and what it has previously encountered. This shift means that the value of what technical writers produce will increase. Their employers can use it to "train" MT systems to produce better results, but doing so will require them to systematize how they capture and store text, if they have not already done so.

At a more practical level, writers can assist the translation task by ensuring that they use terminology consistently, write clear and unambiguous sentences that are self-contained and do not rely on context, and limit complexity. These tasks also improve the usability of texts for many users, so they deliver benefits beyond translation. Careful writing that keeps the needs of MT and human translators in mind is not difficult, but delivers considerable savings when text is translated into many languages.

Will MT replace human translators?

Professional translators frequently worry that MT will replace them with a cheap technology that delivers inferior results. Although most enterprises would doubtlessly like to eliminate the cost of human translation, the good news is that MT is not doing away with translators. Although it is displacing some types of human translation, enterprises are increasing their outlay on professional translation even as they spend more on MT. Although not all translators want to produce PEMT, demand for it will more than double from 2016 to 2019, and this represents a growth opportunity for those linguists who can produce it efficiently.

In other words, MT is increasing the overall translation market – and all segments within it – rather than taking away from translators in a zero-sum game. Although some translators will lose particular work to MT, the overall trend for translators is positive. The role of translators will change, with increasing emphasis on their ability to combine the result of multiple technologies with their own linguistic abilities. These technologies are making translators more productive and allowing those who embrace them to increase their effective hourly pay even as they decrease per-word rates. In this regard, translators are like professionals in other industries – such as accountants – who have seen the value of their skills increase even as automation has eliminated many tasks they formerly carried out.

Conclusion

Machine translation is just the tip of the iceberg. Other technologies – such as automatic content enrichment, sentiment analysis, and improved human-oriented translation tools – are also transforming translation as a profession. The good news is that they make translation more efficient and create growth opportunities for all parties.

Content creators can dramatically increase the amount of content they translate or the languages they support. LSPs that embrace language technology are finding that doing so enables substantial increases in revenue. Technical communicators discover that their content becomes more useful and can help facilitate global content strategies. Translators may see dramatic changes in their work, but those who take a positive approach to technology find that it makes their work more efficient and valuable, enabling them to compensate for falling per-word prices. Despite fears that MT would destroy translation, it appears instead to be a rising tide that lifts all boats, at least those not anchored to how things were done in the past.

This article is based on proprietary research, which is detailed in the following CSA Research: "MT’s Journey to the Enterprise" (May 2016), "Post-Editing Goes Mainstream" (June 2016), and "Fast-Growing LSPs Turn to Machine Translation" (June 2016).