Machine translation (MT) is a hot topic, discussed at every industry conference, in print, in online forums and anytime two or more localization practitioners get together. Oddly, while it is a mainstream topic, it is not a mainstream reality. A lot more people are talking about doing it than there are people actually doing it. There are a number of reasons for this, but four drivers really stand out.
Vendors do not know how to sell MT
While there are a number of good MT providers and a number of good language service provider (LSPs), an alarmingly small number deliver on both. That is not to say that every LSP has to have an MT offering of its own, but every LSP should have a few trusted MT partners whose services they can resell to the client. LSPs should have an opinion, a viewpoint on MT, a strategy that every salesperson can articulate and an operations team who knows how to integrate it into the workflow. LSPs should be able to say to a customer, “Right, you want to do MT in 12 languages; we can do that. We will align the right technology partners and the right strategy for each language, manage the training and implementation, and we will bundle those services for you.” A good MT offering will build pricing into the current word model; it should not be a completely different pricing discussion.
Clients do not know how to buy MT
There is a myriad of considerations — proprietary or open source; statistical, rules-based or hybrid; post-edited or not; pre- and post-processing; what languages; what content; what return on investment. Since LSPs are not there with the answer, this means clients have to do a lot of research, traverse a lot of dark alleys, and in the end, hope they have arrived at the right conclusion. It’s an iterative process potentially littered with costly mistakes. Often, it is simply too daunting and is not pursued. Common Sense Advisory’s maturity model tells us that only 40% of localization departments with more than ten years of management are engaged in MT. The numbers are much smaller for younger operations — testaments to the difficulty clients have in launching MT.
Translators do not want MT
The same scenario has played out in a number of industries. Translators are afraid that their livelihood will either evaporate or evolve into something they do not enjoy. Continually pushed on productivity, they view MT as one way to try to squeeze more out of them. Beyond that, there is the sense that MT removes the art from the craft. That sentiment is not without merit and drives an inflection point for translators. Ultimately, translators will need to choose between two paths: high-end transcreation work or continuous, high volume post-editing, be it from a translation memory, MT or a hybrid of the two. There will not be a middle ground in the future.
The industry has undersold MT
Interwoven with the first three drivers is the fact that MT quality, just a few years ago, was only good enough for gisting and secondtier support content. The low quality meant selling it at a low cost. The conundrum is that with a reputation of low quality and low cost, moving up the value chain is incredibly difficult. An avalanche of data indicating that MT quality has vastly improved hardly overcomes the ingrained negative perception. The pricing only helps to solidify that perception because anyone with a little business savvy knows you do not sell a good product cheap.
What is next for MT?
It’s not all doom and gloom. There are examples of businesses, ideas and products that have overcome entering the marketing at the wrong end. It’s going to require LSPs getting a lot more savvy about how they market MT. They are going to have to engage the core of their supply chain — the translator — to understand their needs and concerns and set an MT strategy that benefits them both. With that done and with the good work already being done by some LSPs and professional industry organizations such as TAUS/TDA, clients can be reengaged with a strategy that is simple to procure. If you cannot make it simple to procure, no matter how good your offering is, it will never ascend out of nichedom.