ATC blog

  • Home
    Home This is where you can find all the blog posts throughout the site.
  • Categories
    Categories Displays a list of categories from this blog.
  • Tags
    Tags Displays a list of tags that have been used in the blog.
  • Bloggers
    Bloggers Search for your favorite blogger from this site.
  • Team Blogs
    Team Blogs Find your favorite team blogs here.
  • Login
    Login Login form

Posted by on in ATC Blog

It is no secret that machine translation (MT) technologies are becoming an increasingly widely implemented feature in online messaging used by customer service departments of transnational companies, and in b2b communications.
At their core, live MT systems usually operate based on a trained machine translation memory (MTM) in a way very similar to the use of TMs in CAT tools. Upon translating the customer’s query based on its memory, the machine offers the employee a suitable answer to use in conversation.
While the main purpose of such solutions is clear and their benefits include hugely reduced waiting time for the customer and a significant cost saving opportunity for the company, system’s implementation and maintenance are quite labour-intensive. Instead of hiring a language proficient agent, the company will still have to train an employee to use the software. Also, the MTM will need to be constantly updated and synchronized. It is very important to keep its contents up to date and reliable since the user will more likely be unable to read the machine-translated text.

Taking into account the wide variety of customer queries, specific terminology, slang, abbreviations and other language quirks, the system could easily drive the online conversation off-course and even affect the customer satisfaction (CSAT) level. For example, a Russian word «Хорошо» can be translated as “Well” in English (as in “You’ve done well”), but is often used similar to “OK” in conversational speach. If a Russian customer contacts an English-speaking customer service rep. who asks them to wait for a moment or two to check something, the machine might translate customer’s «Хорошо» as “well” (instead of “Ok”), making the customer look impatient or irritated. This is one of the many actual examples of small miscommunication incidents that might happen while using live MT software. However, if you compare their language barrier breaking capabilities as well as the Service Level Agreement (SLA) benefits for the companies utilizing them, it becomes clear that the overall value of MT engines is enormous.

A range of such solutions are employed by different companies and implemented into various live communication tools (e.g. liveperson , or Geofluent). There are also solutions more oriented on b2b communication, e.g. https://www.chatlingual.com/

On the other hand, we see an increasing demand for online services where live translation is performed by remote human translators rather then MT engines. Platforms such as SpeakUs or Cloud Interpreter focus their effort on facilitating “live interpretation from any location via browser or application” with the help of professional interpreters from all over the world.

In one way or another, technologies helps us all stay connected. And we hope that with the aid of NMT (Neural Machine Translation), live MT systems will only improve CSAT.

Hits: 569
Rate this blog entry:

Treating MTPE as just a brief manual revision of the automatically generated translation before the end user delivery would never let us achieve the results mentioned by eBay’s Senior Director of Machine Translation and Geo Expansion, Hassan Sawaf: “As we’ve rolled out our MT capabilities, and even before a lot of the education and outreach we plan to do, we’ve quickly increased the number of Russian users we see using these features by 50 %”.

For Russian, which is an inflectional language with a complex structure and high morphological demands, every single MT segment has to undergo a multi-stage processing sequence with several linguists working on it in turns.

- Together with the client we developed a comprehensive guideline on the language specific conventions of MTPE, and made sure that each member of the team was adhering to it. This guide was continuously being updated.

- Each MT-segment was processed with a usual 4-step TEP+QA worfkflow, modified for the project: MT-editor/2nd editor/proofreader/QA specialist performing automated checks.

- Resources for each of these steps were tested through a special procedure, adapted for the project specifics. (The content we worked on was meant not for human users, but for the MT engine’s training.)

So this ride was even more complex than a standard localization cycle. Then why bother complicating time-tested processes, get paid less, and not just “translate from scratch”, as usual? The answer is, MT is not as black as it is sometimes painted. Judging from the experience with such major accounts as eBay, Cisco and Dell, we do believe that MT is good. But it is certainly not yet capable of replacing HT: if our goal is the client’s satisfaction, there’s always a job to do for human experts. 

We especially doubt that any MT engine can be trained well enough to produce a near final-quality translation for Russian and other complex languages. As recent article by Memsource states, “Russian, Polish and Korean have lower MT leverage rates, below 40% or even 20% fuzzy matches and 5% complete matches.”

Back to the eBay case, we think the 50% increase in the number of Russian users was achieved mostly because the content was translated. And although MT is not a universal remedy, implementing it played the key role in the success of this particular case.

In many other cases, it’s better to have no translation than a poor one (which raw MT output usually constitutes).

For a lower standard often referred to as “fit for purpose”, light PE may suffice, which aims to make the MT output “simply understandable”. However, in our 5+ years of MTPE practice we’ve never faced an actual project with Light MTPE demands. On the contrary, those of our clients who utilise MT, tend to present some of the highest quality expectations. This is probably because they’re putting so much effort into MT deployment, including engine training, output evaluation, analytics, statistics, not to mention the actual PE work for each language involved. Consequently, highest quality is expected.

Hits: 1158
Rate this blog entry:

EUATC LOGO

Newsletter

Email address:


Follow Us