How to Create a Pipeline in Rainbow

From Okapi Framework
Jump to navigation Jump to search

You can easily create and execute pipelines in Rainbow. In this example you will be creating a pipeline to automatically translate any source document in a format supported by Rainbow.

1. Start Rainbow.

2. Drop a source document, for example an HTML file, in the Input List 1 tab. Specify the source and target languages and well as encoding information as needed, in the Languages and Encodings tab.

3. Select Utilities > Edit / Execute Pipeline.

Pipeline dialog in Rainbow (Windows 7)

Creating a new pipeline is done by adding the steps it uses, in the order they need to be executed.

In this example you need to perform the following actions for each input document:

  1. Extract the translatable from the original format.
  2. Translate all extracted text using an MT system.
  3. Merge back the extracted text into its original format.

Each of these actions can be performed by a corresponding step in a pipeline like this:

= Raw Document to Filter Events Step
+ Leveraging Step
+ Filter Events to Raw Document Step

4. Click Add Step, select the Raw Document to Filter Events Step, click OK.

This step applies the filter configuration associated with the input document and decompose the content into filter events, which carry the translatable parts. There is no option to set: all the information needed is set with each the input document.

5. Click Add Step, select the Leveraging Step, click OK.

This step recieves the filter events and for each text unit, it tries to find a translation for the source text using a specified translation resource. The translation resource can be an MT engine, a TM, etc. In this example we can use one of the easier MT system to access: Google MT. For more information on translation resources see the "Connectors" page.

Make sure the option Leverage the text units with existing translations is set, and select Google Translate Services.

Translation proposals coming from MT systems have a score of 95%, so you must also make sure the option Leverage only if the match is equal or above this score is set to 95 or lower.

Make sure the option Fill the target with the leveraged translation is also set so the translated text is copied into the text unit and can be output.

6. Click Add Step, select the Filter Events to Raw Document Step, click OK.

This last step simply re-creates the original file with the translated text. All the information is already set with each document.

At this point the pipeline is ready to be used. You can save the pipeline and its options into a file using Save As. You will be able to re-load that pipeline using Load at any time.

If you click Close: the dialog box closed without running the pipeline. The pipeline will be preserved for the next time you open the dialog box. The pipeline in the dialog box is also saved automatically with the Rainbow project when you save the project.

If you click Cancel: the dialog box is closed without saving your modifications.

7. For this example click Execute to run the pipeline. Rainbow will take whatever input files you have listed and run them through the pipeline.

When the process is finished you should have an output file where all translatable text has been translated using Google Translate.