101

1. What is MT?

Read and check the following content

Determine if the following statements are true or false

  • Machine translation helps you translate more content in less time which will be more profitable for you.
  • Rule based systems are based on millions of words.
  • Machine translation is that translation you can create with the assistance of a computer.
  • Statistical systems can be easily used to translate from source into target and target into source.
  • KantanMT will help me by creating the engines for me and putting them into my hands when ready.
  • Storing my data on the cloud means that everyone can have access to them.

2. What is KantanMT?

Read and check the following content

Now have a look at the following video to get an overview on what the KantanMT Platform looks like. It will help you get started with your tasks.

 

 

Where should I go to...

  • ...create groups of data?
    Templates
  • ...manage and activate my profiles?
    Clients Profiles
  • ... build an engine?
    Training Data
  • ... link your engine to your CAT tools?
    API
  • ... translate documents
    Client Files

3. Creating a MT engine in KantanMT

Creating Client Profiles With KantanMT

This video will show you how to create a profile in KantanMT:

 

 

Adding Training Data to a Profile & Build a KantanMT Engine

This video will show you how to upload training data and how to build your first KantanMT engine.

 

 

How to build an engine and put it to work

  • Activate profile
  • Go to your Training Data Tab
  • Upload Training Data
  • Click Build
  • Go to my Jobs to check the task
  • Go to Client Files
  • Upload Files to translate
  • Click Translate
  • Go to my Jobs to check the task and download the output

4. Types of Training Data

Read and check the following content

Which of the following scenarios would result in a properly built engine?

  • 3 tmx, 2 PDFs, 1 glossary.xls and 2 stock data
  • 1 tmx, 1 source.mono.utf8.trg, 1 glossary.tbx and 1 stock data
  • 4 tmx, 3 docx, 1 Glossary.tbx and 2 stock data
  • 1 source.utf8.src, 1 source.utf8.trg, 1 terminology.xlsx, 1 source.utf8.trg.mono and 3 stock data
  • 2 tmx, 1 source.utf8.trg, 1 Glossary.xlsx and 2 stock data

5. Steps to Build an Engine

Read and check the following content

Please fill in the gaps

I want to create a new engine from scratch. First I will need to add my and . Once it's finished I will   the scores and  the needs of my engine. Now I will add my  and  , and train the engine again. I will follow the same process and when it's done, I will check my . If they are low, I can add  and  again to boost the scores and have better results when I .

With these three iterations, following our three steps, we will manage to create an engine that will be ready for !

6. Machine Translation Scores

Read and check the following content

Which ideal figure corresponds to which score?

  • F-Measure Score
    > 70%
  • TER Score
    < 40%
  • BLEU Score
    > 60%
  • Monolingual Word Count
    2 million
  • Unique Word Count
    1 million
  • Bilingual Source Word Count
    > 5 million

7. Build Analytics

Using BuildAnalytics™

Understanding BuildAnalytics™ Scores

 

Where can I find the following reports?

  • Untranslated content from my data
    Gap Analysis
  • Invalid segments from my bilingual data
    Rejects Report
  • Backlog of my engine
    Timeline
  • The precision of my data
    F-Measure Score Report
  • Quality of the target language from my data
    BLEU Score Report
  • Post-editing effort
    TER Score Report