Evaluating Slate™ Desktop

Evaluating Slate™ Desktop – If you think evaluating MT is complicated, there’s good news. Three simple steps make it easy and you will discover the immediate if your MT system benefits you. In our first newsletter, A Proper Introduction to Slate™ Desktop, I introduced you to the personalized translation engine™. This newsletter introduces you to our 60-day risk-free trial to evaluate Slate™ Desktop.

In our first newsletter, A Proper Introduction to Slate™ Desktop, I introduced you to the personalized translation engine™. This newsletter introduces you to our 60-day risk-free trial to evaluate Slate™ Desktop. Traditional evaluation relies on a patchwork of machine translation post-editing (MT-PE) skills dating back to the early 1960’s. Those best practices were developed to compensate for MT’s failure to create translations that you deem correct. Sadly, those best practices have not kept pace with the technology.

Slate™ Desktop takes a holistic approach to evaluation that eliminates MT-PE altogether.

Step one – Create an engine

Your first step creates an engine to evaluate. While Slate™ Desktop creates your engine, it learns what you expect and how to mimic your expectations. This is the essence of personalization and it greatly simplifies your evaluation.

Slate™ Desktop creates an “Engine Summary” report for each engine. The report’s Quality quotient estimates the engine’s capacity to mimic you. This screenshot shows the engine replicated 37.5% of this translator’s 2,381 TUs. You can expect similar results with daily projects so long as you continue working in the same field.

The Quality quotient is your first evaluation milestone:

  • Less than 15% – Your engine will perform similarly to Google and Microsoft. I recommend you stop your evaluation and request a refund unless you need the confidentiality benefits.
  • 15% to 25% – You probably work with a difficult language pairs or with creative content like gaming, literature or marketing. I recommend you continue your evaluation and carefully scrutinize the results for your needs.
  • 25% to 45% – You’re among the most common Slate™ Desktop user group. Continue your evaluations and start measuring the changes in your capacity.
Step 2 – Review the evaluation set

Continuing to the next step, you should review the evaluation set that generated the Quality quotient. They indicate how your engine will perform with new projects. You’ll find the set in a spreadsheet (instructions on our support site). You can sort and group the rows to quickly find the exact and near replica TUs.

Final step – Use your engine

In the final step, you work with your engine in a project. Let’s calculate real numbers for a project with 10,000 words, 625 segments and the Engine Summary above. Without Slate™ Desktop, you normally work a project from beginning to end at 250 words-per-hour for 8 hours a day, your capacity is 10,000 words over 5 days.

We recommend you modify your work to include these sub-steps to learn about your engine:

  1. Configure a CAT project to use your new engine.
  2. Note your start time.
  3. Work through all segments to generate and proofread your engine’s suggestions:
    1. If you judge a suggestion as correct, lock the segment.
    2. Otherwise, don’t change the segment and move to the next one.
  4. Note your end time. You proofed 625 segments in 2 hours.
  5. Count the locked segments. This project locked 127 segments with 1,921 words (20%).
  6. Compare locked with the Engine’s Quality quotient. A 10-15% drop is normal.
  7. Resume working on the unlocked segments. Use your judgement to fix the Slate™ Desktop suggestions or re-translate. This project had an additional 120 segments with only 1 to 3 word change.

With Slate™ Desktop, you’re still working at 250 words-per-hour. The difference is your engine finished 20% of the work for you and reduced another 20% of your work. Overall, it generated 245 segments that required 0 to 3 word changes. As a result, you finished evaluating your personalized engine and you finished a project one day faster.


When you’re done, here’s your time invested in the three evaluation steps:

  1. Create your engine: 30 minutes of your attention plus Slate™ Desktop worked overnight.
  2. Review evaluation set: 30 minutes of your attention.
  3. Use your engine: 2 hours of your attention and potentially saves you a whole day’s work.

With the 60-day trial, our money-back guarantee, and 3 hours invested in evaluation, you know how Slate™ Desktop will work for you.

What are you waiting for? Evaluating your personalized translation engine is easy!

Yours truly,
Tom Hoar
Managing Partner
Slate Rocks LLC