Following KantanMT’s announcement of the roll out of the much-anticipated KantanLQR™ platform to all its Partners worldwide, Louise Irwin from Digital Marketing Team caught up Louise Faherty, Project Manager, Professional Services, KantanMT to talk about the features, benefits and the impetus behind creating the tool.
Louise Irwin (LI): Before we begin, for our readers who missed our press release on KantanLQR, could you give us a gist of what it is all about?
Louise Faherty (LF): KantanLQR is revolutionary quality review tool that automates the language review process, which in turn will create MT engines of a higher calibre and decrease project turnaround time by allowing Project Managers (PMs) to create production-ready engines faster.
LI: Brilliant! That sounds pretty revolutionary, alright. What was the impetus behind creating this tool? Don’t we at KantanMT already have a system in place for Language Quality Review (LQR)?
LF: Yes, we do, but if you ask any LSP or enterprise that works with MT, they will tell you that the LQR process can be a very tedious, long-drawn and often frustrating process.
LQR, as you know is an absolutely vital step to creating high quality KantanMT engines. Once the engine has been built, we run segments to be translated. The source and translated segments are then put in an MS Excel sheet and sent to translators for quality check. PMs have their own parameters or Key Performance Indicators (KPIs) that they want to review, and they ask the translators to score the translated segments based on these KPIs.
Once the translators send their feedback, the MT engineers use this to improve the quality of the segments. This is an iterative process and in most projects you would end up being inundated with Excel sheets! Not to mention, the manual nature of traditional LQR makes it a long-drawn process and can leave room for a lot of human errors.
So, to go back to your initial question, what was the impetus behind KantanLQR: Simply put, we wanted to further automate the Custom Machine Translation (CMT) process.
One of the oldest clients we work with is an eCommerce giant, and they have a very high volume of content to be translated in more than 14 language combinations. While working in projects with this client, we saw that a high-calibre engine would take about 6 weeks to build, out of which 4 weeks would be dedicated to LQR. We started looking at ways to shorten the engine building time by 2 weeks. We realised that if we automate the LQR process and make it more efficient, we could provide an even more effective CMT solution to our client. That was the impetus and the driving force behind developing KantanLQR.
LI: When you say KantanLQR automates the language quality review process, do you mean you are taking the human element out of the process?
LF: No! Not at all! Our reviewers are vital to the process. KantanLQR is simply meant to provide a platform where Project Managers can benefit from a more dynamic and productive method of measuring translation quality that offers formalised metrics. This will drive a deeper understanding of how an engine will perform in production.
PMs can chose from built-in KPIs (translation adequacy, fluency, overall quality, style, syntax and grammar).
Translators score the segments on these KPIs and can also leave a comment. The scores are automatically reflected on the PM’s dashboard and is updated in real-time. Depending on the progress of the project, which is also updated in real-time, a PM might decide that the translation quality is high enough and the engines are production ready, even before the linguists have reviewed all the segments.
LI: What are the key performance indicators all about?
LF: KantanLQR has been designed to be extremely easy to use. We have stuck to the same clear, quick start focus of KantanMT. KantanLQR also follows the same methodology as a standard human quality review, so users will require minimal training and can get started very quickly.
Once a user signs up on the platform, they can create and customise language review projects by selecting a set of built-in KPIs including overall quality, translation adequacy and fluency, and style, syntax and grammar, which are similar to the parameters set out in the Multidimensional Quality Metrics framework (MQM).
What’s better is that a PM can enter their own custom criteria, if they need. This makes the process streamlined, resulting in faster review process and quicker project turnaround time.
LI: Is there anything else our users should know about KantanLQR?
LF: Speaking from a project manager’s background, KantanLQR is going to take the guesswork out of human linguistic quality review.
It will give us a unique, real time insight into the progress of a review, which can have a real impact on the whole project. For example, if a project scores very high, we could decide to launch an engine early. On the other hand, if there is a problem with an engine we will be able to see this in real time, and we might decide to stop the review and check the engine again.
KantanLQR also allows use crowdsourcing – there is no limit to the number of reviewers one can add to a project, meaning a faster turnaround. KantanLQR will automatically present un-reviewed segments to each reviewer and there is real time reporting on reviewer productivity too. Reviewer bias will be absorbed by using a number of reviewers which makes your review fairer. With all these possibilities we are so excited to hear how the KantanMT community uses the product- we are sure they’ll discover even better methods too!
LI: Louise, thank you so much for taking the time out today to tell our readers more about the features of KantanLQR™.
To all our readers, we would like to remind once again that KantanLQR is now available to all our partners. To know more about how to begin using the tool, please mail info@kantanmt.com.