Machine Translation as a service?

Should LSCs offer machine translation as a service?

This is a question being pondered by company managers throughout the language industry as many companies are still grappling with their strategy around machine translation. Companies of many sizes struggle with moving from concept to production using machine translation. That many firms are expending the effort is an encouraging sign for the industry, which reflects a healthy skepticism towards the technology.

There is much at stake: stand on the sidelines like a veritable buggy whip maker and watch the automobile steal your livelihood or get behind the wheel and steer your own destiny. Most firms are beyond the philosophical considerations of “Should we use MT?” and are contending more with the “How do we use MT?”. One question that arises consistently (and which I find puzzling) is, “Should we sell machine translation as a service?” I strongly recommend that the majority of LSCs should forego this question altogether and here’s why…

Why rush to sell a commodity?

Thanks to the biggest players in the tech world, nominally effective MT has quickly become ubiquitous. Anyone with the slightest curiosity can setup a Microsoft Translator, Google Translate, or Amazon Translation account and begin translating content themselves for free. Sure, there is a subset of customers who are too lazy or disinterested in doing it themselves, but the value that an LSC can add to selling raw MT is so minimal that the justifiable market isn’t even worth their effort.

You don’t make it, so why sell it?

I would argue that if you a company that has its own MT solution then it’s a different value proposition, but even then, you are competing against the Big Three tech giants, so your solution better be world beating and offer something theirs can’t. I would say that one area that could become a separate service is the development of specialized engines. Firms that can develop the infrastructure and processes to train generic engines to produce customized content for specific customers or industrial sectors will differentiate themselves. However, I predict that given the relatively low barrier to entry to this type of service that the language industry will soon see a glut of MT Services providers.

Why sell a service that cannibalizes your other services?

Some LSCs are attempting the “tiered service” approach, which offers customers differing levels of quality. Except for large-scale translation for information purposes only, I don’t see this as an effective strategy. It’s rare that a customer will even understand or have identified the quality requirements of their intended audience, nor will they take the risk of offering their customers “less than perfect” translation. All this strategy does is enable translation buyers to self-qualify as those who care about what they provide their customers and those who don’t care quite as much (or have limited budgets and must simply scrape by). The hard truth for LSCs is that if they go down this path, they are simply cannibalizing their core value-added services in favor of the quicker and cheaper sale.

That “ethics” thing

Many LCPs believe they must offer machine translation as a service so that clients are fully aware of what they are buying. While I understand this logic, I refuse to accept that using MT as part of production practices is unethical if not disclosed to the client. What is unethical is using MT without adding any value in the process and selling raw MT as if it were human quality translation. That is patently unethical and occurs consistently within the industry itself with unscrupulous freelancers selling raw MT as a final product to unsuspecting LSCs. If your firm chooses to use machine translation as a service to improve turnaround and manage costs, that’s precisely why automation exists in business. Using tools effectively and appropriately is the hallmark of all professional practitioners regardless of vocation–this is no different for translation. If you are having minor surgery, do you quibble with the surgeon if he will be using sutures or medical adhesive? No, you trust her judgement and allow her to make her own professional decision.

As an industry, we must have the confidence of experts and choose our toolsets as we deem necessary. Not all content should be run through translation memory, for example. And, similarly, not all content can and should be run through an MT engine. There are content types that cannot be reliably translated by an engine and translation professionals should know when to engage a specific class of tools or not for best effect.

I’m often asked by LSPs, “but what if a customer asks if we use MT?”. I’m a fan of telling the truth and trusting our clients–that, if well informed, will understand and trust our production choices. If asked, simply tell your client yes, and here’s why…Your discussion will more than likely gravitate from discussions about quality to re-negotiating rates. This is another reason I don’t believe the industry should rush to put MT front and center in process discussions with customers.

Erosion is a natural process…therefore, don’t encourage it

Price erosion is real and even before the advent of practicable MT in production, translation rates have been steadily eroding for the last 25 years–thanks to globalization and technology. Let’s not make the same mistake with machine translation that the industry made following the widespread use of translation memory tools. In the late nineties, the technology arms race amongst LSPs began reaching fever pitch with the advent of CAT tools. With little regard or understanding of markets, early adopters of translation memory began touting their use of these tools and began price wars by unilaterally offering TM discounts. Most clients cared little about the efficiency benefits of using translation memory tools. They cared mostly about the cost benefits and the resulting lower effective per-word rates they were paying.

Given the nascent status of MT, competitive advantage exists for those LSPs who have committed the time and resources to developing a strategy and resources needed to make MT operational within their production process. This is the value proposition for your customers, and they should benefit from your firm’s greater efficiency, but by how much is where the free hand of the market should come into play.

So, I beseech those LSPs who are toying with the idea of utilizing MT to become expert at using the technology first and worry less about how to market it. First deploy it, make it your own, establish your differentiation in the market by using it, and enjoy the benefits of that position for as long as they may last.