January 2016 Hub Release Notes RRS feed

  • General discussion

  • We have just completed the January 2016 release of the Hub.


    Here are the updates:

    • Dictionary only training: You can now train a custom translation system when you just have a dictionary and no other parallel documents. There is no minimum size to that dictionary. It should have at least one entry. Simply upload the dictionary, which is an Excel file with the language ID as column header, include it in your training set, and hit train. The training completes very quickly, and you are ready to deploy.
    • Training with 1000 parallel sentences only: You can now train a custom system with only 1000 parallel sentences. Use 500 sentences for the tuning set and 500 sentences in the test set. The Hub will build a system based on Microsoft models, and will tune the models to your tuning set, giving you a better adjusted system than the generic translation system. You can use in-domain target language documents as part of this training as well. The 1000 sentences must be unique and pass the Hub’s data filtering.
    • Both of the above training types work only for language pairs with Microsoft models. There is more background and information about these changes on our blog.
    • Easily scroll through your systems: You can now adjust how many systems you want to see at once when browsing. You can also download the overview of your systems and inspect them using Excel.
    • French-Canadian: Now available as a language choice. When using Microsoft models, a French-Canadian systems will use the regular “French” Microsoft models, but the availability of French Canadian (FR-CA) in the Hub allows you to differentiate in your dictionaries and your own training material.


    Let us know what you think of these changes, and any questions you have about them.

    Tuesday, February 2, 2016 12:20 AM

All replies

  • Thank Chris. I have found when attempting to start training with Dictionary only (for English to Japanese), it gives me a warning that says "Cannot submit Lambda Or DictionaryOnly training as Microsoft models do not exist for the selected language pair". It then fails without any real debugging information. Is this correct?
    Monday, February 8, 2016 11:43 AM
  • Hello Simon,

    trainings with only a dictionary are currently failing for systems TO Japanese and systems TO Malay. We expect to have this addressed on or before Thursday February 11.

    Other types of training and other languages are not affected. Or let us know if you see something strange.

    Chris Wendt
    Microsoft Translator

    Monday, February 8, 2016 4:36 PM
  • Hi Chris - do you know if this was updated? I managed to get a deployment (which I am waiting a response from the Translation Team on) by finding a bunch of translated sentences and used them for training, even though we only need to Dictionary.
    Friday, February 12, 2016 4:35 AM
  • Hi Simon,

    We have the fix ready to go, but we are experiencing an issue with our production setup. We will rollout this fix by Tuesday (16th Feb). Apologies for additional delay. 

    Thank you,

    MT Hub Support team

    Friday, February 12, 2016 8:38 PM
  • We have fixed issue with Japanese-English/English-Japanese dictionary only training. Please give it a try. 

    Thank you

    MT Hub Support Team

    Monday, February 22, 2016 4:28 PM
  • It still won't let me proceed and I have tried logging out and back in and creating a new system.

    hub.microsofttranslator.com says:
    Please add/select testing, tuning, training dataset documents before conducting the training.

    Tuesday, February 23, 2016 3:41 AM
  • Hi Simon,

    We are able to create Japanese English dictionary only training. We are investigating to see why you are seeing message to add more data for dictionary only training. We will provide an update asap.

    Thank you

    Tuesday, February 23, 2016 4:25 PM
  • Hi Simon,

    We are not supporting dictionary only feature for English (United Kingdom)(en-gb). It is available for English (en/en-us). You can create a new training with English (en/en-us). Please rename your en-gb files to include suffix en/en-us and upload to your files to new training. You should be able to conduct dictionary only training. You would get same translation experience with this new training (in this case English to Japanese). Please let us know if you need any additional details.

    Thank you

     MT Hub Support team   

    Tuesday, February 23, 2016 5:40 PM