Molly Ancello: Hi everyone, I’m Molly Ancello, Marketing Coordinator at Advanced Language Translation. I am here today with company President Scott Bass.

Scott Bass: Hi Molly

MA: Hi Scott. So today I’m going to talk to Scott about Automated Translation or Machine Translation as we call it in the industry. So just to note a few things: we will use the terms Automated Translation and Machine Translation interchangeably throughout this recording and the other thing is that when we mention ALT, we are referring to our own company Advanced Language Translation.

So who is this podcast for? It’s for those of you that are at a company that already translates, or for those of you that are maybe new to translation. And for those of you that haven’t explored Automated Translation, this is a good learning opportunity.

So Scott why don’t we start off with a brief background of Automated Translation and why don’t you tell us a little bit about where the technology stands today.

SB: Sure. Machine Translation as I’ll refer to it is actually pretty old.It started in the 1940’s following WWII as the Cold War started heating up. And a lot of the funding was driven by the US government to develop the technology for doing Automatic Translation of intelligence material from English to Russian. By the 1970’s it had gotten to the point where it could be commercialized. So it was being used by some companies to translate a high volume of material. But then by the 1990’s and the rise of the World Wide Web,we really saw it become more publically used in the form of BabelFish for example, where you could translate pages on the web that you wanted to get the gist of. So then flash-forward to the 2000’s and the introduction of Google Translate. That was a big step forward for the technology because it was making use of what is called statistical machine translation, that ALT is using.

MA: Ok, so when did ALT actually embrace Automated Translation? And why? Why not stick with all human translators?

SB: Well quite frankly its just a matter of how business evolves and we’re always looking for faster and cheaper ways to get the same amount of work done. Translation is no different than any other industry, and automation is always the answer. ALT has been watching the evolution and progress of machine translation for the last twenty years, always with interest to see whether it is something that is going to help us…is it a threat to how we do our work? Is it a tool we can use? And then again with the advent of better technology in the form of statistical machine translation it got to the point where we said hey this is usable stuff, we can actually work with it. But the technology isn’t magic; we have to build and customize engines to make them effective. The process requires lots of care and feeding.

MA: Ok so Automated Translation isn’t magic. You mentioned that a translation engine actually has to be built… and how long does that usually take, and what does it entail?

SB: Well our technology partner, Asia Online builds what are called foundation engines. These are general engines setup for specific language combinations and general subject terminology. Engines require lots of high quality data in order to generate translation of useable quality; so starting with a good foundation engine, we typically need about six weeks to prepare a customized engine.

MA: Then once an engine is built what exactly happens with a client’s content?

SB: All machine-translated content is reviewed and improved by human editors. This process is similar to the traditional all-human process, in which a draft translation created by a human translator is reviewed and edited by a second translator. The only difference with Machine Translation is that the draft is prepared by a machine. The post-edited content is then fed back to the translation engine to improve it for future use. This process is called tuning and it can significantly improve the quality of machine generated translation.

MA: So machine translation, the process parallels that of traditional human translation. In both cases there is an initial translation and editing that follows.

SB: Exactly.

MA: And the benefit of Automated Translation is the tuning process that you mentioned, where the errors that the human editor corrected are actually “saved” so you are essentially training the engine to perform in a certain way.

SB: Right.

MA: So basically with each additional project the engine becomes better and more precise.

So what percentage would you say of ALT’s clients is currently using Automated Translation?

SB: Well percentage-wise it is a small number in terms of the number of clients. But in terms of the amount of translated content it’s significant. As much as 33% of our work volume is passed through machine translation. On average customers save 30% in translation costs and turnaround is at least 50% faster using Machine Translation instead of using an all-human process.

MA: Wow. So huge savings on both budget and turnaround.

So now tell me, what is the standard profile for a company using Automated Translation?

SB: There is not one specific profile that a company must fit to benefit from machine translation. But we see the best results with companies that require a high volume of translation. That would be approximately two languages that need about 500,000 words each per year. Technical subject matter such as user or operator manuals for machines or software can benefit greatly from this type of approach as well. But, most importantly the companies must have a lot of existing translation, whose content is already stored in translation memory is a good basis to start from. This is the most efficient way to use good quality data to customize the translation engines.

MA: Ok great, so companies should already own their translation memories so that they can apply the translation memories to the engines and build a good foundation from the start, right?

SB: Right.

MA: And if a vendor charges you to have access to your translation memories or claims that you don’t own them, then you should question that because those should be your assets not your vendor’s.

So back to the ideal company profile for automated translation: businesses that translate more often with higher volume projects into more than two languages are ideal for machine translation. So are you saying Scott that a company with smaller amounts of material, or with more sporadic translation projects may be a better candidate for human translation?

SB: On the whole, yes. The amount of preparation and data it takes to customize an engine may far outweigh the benefits if the company is only doing small and sporadic projects. Companies with sporadic need and smaller translation volumes should not expect their projects to be a good fit for machine translation. They are however, perfect for human translation.

MA: So what other factors should our listeners consider to determine whether their company is right for Automated Translation?

SB: First of all, translation must be an important tool in a company’s marketing strategy. In other words, international markets must represent a significant growth opportunity to a company. Otherwise, translation itself will not be valued and will not be worth the time, effort and cost. Companies that are already engaged in international markets and whose sales will benefit from translating more content — such as customer support documents — those companies will benefit from automated translation.

MA: Now let’s say our listeners have decided that they actually fit the ideal company profile for automated translation…what is the right time to introduce it?

SB: Before answering that question, it’s important to address a company’s ability to even measure how much translation they are doing and projecting how much they will do in the future. Companies that take translation seriously will carefully track how much translation they carry out each year and how much those efforts cost. Based on that information, they should be able to project how much translation they will need during an upcoming year. Once they get into the range of about ½ million words per language per year, they should begin to evaluate Automated Translation as a solution.

MA: So that about wraps it up, and now you all have an overview of Automated Translation and a guide to helping you decide whether it’s right for your company. We have included the transcript to this podcast so that you can revisit this information at anytime.

I want to thank Scott for talking with me today and allowing me to pick his brain about Automated Translation. I also want to thank you, our listener for your time and interest in this subject. Hopefully this discussion has helped you realize that your business is either right for automated translation or human translation. As always we are happy to answer any questions that you have.

Thank you.

SB: Thanks a lot Molly.