Training and Managing Memories

Memories are the core of Lilt. You need them to create projects, train your custom MT engine and make translation requests. Each memory has a unique identifier which you can look up by opening your memories page in the Lilt app in your browser. Or you can retrieve all your memories with this simple API request:

$ curl

If you haven't created a custom Memory yet, you can easily create one:

$ curl -k --data "srclang=de&trglang=en&name=MYMEMORY"

If you created a custom Memory via the web app, please verify that you can access the Memory. In case you haven't created the Memory with your master (API) account you may have to explicitly share that Memory by clicking on the share icon.

You can also use your default Memory. You have a default Memory for every language pair and it is initialized if you have previously associated a project with it. Your default Memory can be identified by searching for the Memory with the name sl_to_tl where sl is the 2-letter source and tl the 2-letter target language code; e.g. German-to-English would be de_to_en.

Training / Adaptation

With every update to the Memory the custom MT engine is trained and adapted to the content in the Memory.

  • Translation memories -- For the best results, upload translation memory files to give the MT engine a head start. The MT is trained instantly without long waiting time.
  • Single segments -- You can add segments to update the Memory. Each new segment helps Lilt adapt to your style and field of specialization and provide better suggestions.

To learn more, see the full API reference.

Still need help? Get in touch!
Last updated on 12th May 2019