File Translation

The File Translation tool provides a quick and easy way to translate a wide variety of different file types, including:

txt, html, htm, xhtml, xml, tmx, docx, pptx, xlsx, pdf, rtf, odp, ods, odt, json, properties, resx, zip, tar, tgz, bmp, jpg, jpeg, png

Note

Systran Translate supports the following json structures: json i18n (see https://webtranslateit.com/en/docs/file_formats/json))

../../_images/JsonUpload_files.png

Note

The translations of uploaded files will be saved for three days. Post-edited files won’t be automatically deleted and will thus be saved indefinitely. In order to delete these files, click Delete next to the file on the File Translation page.

Translating files

To start translating files, set the translation options:

../../_images/PRO.Doc9-ug-tools-11.png

Use the provided drop-down menus to select the Source language and Target language for the translation task:

  • Source language: The language of the original text, prior to translation

  • Target language: The language into which the source text will be translated

  • Use the provided drop-down to set a Profile (a set collection of Translation Options and Resources).

Select the file(s) to translate:

  • Click Upload to open the Upload Files dialog

  • Drag & drop files or click Select Files to select one or more files to translate

Note

The file size limit is 15MB

../../_images/9.3Doc9-ug-tools-12.png

  • Click Upload

By default, all files added will be uploaded if they are not unchecked in the table. Only accepted formats are automatically selected:

../../_images/9.3Doc9-ug-tools-13.png

  • Depending on the file type, specifying the format may be required. For example:

    • Plain text: Text format including .txt, Microsoft Word file types, etc.

    • Bitext: A file containing both source and target language versions of a given text

    • Token form: A file in which the text is tokenized

  • If for any reason you wish to cancel the upload, select the file to cancel then click Remove

  • When uploading is done, the status “Successfully uploaded” will appear:

../../_images/9.3Status_success.png

  • You can also click on Translate all & close to upload and initiate translation of all the files.

  • Your file will appear in the table on the File Translation page displaying all uploaded files

../../_images/9.3Doc9-ug-tools-14.png

Consult this table to check:

  1. The names of uploaded files

    • Click on the small triangle in the Filename column header to sort alphabetically by filename

  2. Status of the translation:

    • queued: request queued for translation

    • dequeued: request no longer queued but not yet sent for translation

    • translating: request sent to translator and in the process of translation

    • translated: translation finished successfully

    • error: translation finished unsuccessfully

  3. Date: How long ago the file was uploaded

    • Click on the triangle in the column header to sort by date

  4. Languages: the source language appears first with the arrow pointing to the target language

    • Click on the triangle in the column header to sort by language (alphabetical order)

  5. Profile selections displays the selectors and the profile used to translate

    • Click on the triangle in the column header to sort by Profile

  6. File size

    • Click on the triangle in the column header to sort by file size

From here, you can also take action on translated files as appropriate, using the provided hot buttons.

  • Edit: Opens the Translation Editor, a tool that allows you to review and manage translations directly on the SYSTRAN Enterprise Server interface

  • Download: Initiates download of the translated file from the server to the local system. The translated file will be saved as such: result_filename.txt

  • Delete: Deletes the translated file and removes it from the server

PDF File Translation

There are four modes to translate PDF files:

  • Precise PDF: This mode parses the text from the PDF. It will provide the best result in terms of text extraction and formatting but requires the PDF to contain extractable text and images will not be processed. The translated result will be a .docx file.

  • OCR: This mode processes Optical Character Recognition on each page of the PDF file and generates a docx for translation. This mode will work for all PDF types, even for scanned PDF and images, but some errors can occur in the character recognition affecting translation, and formatting can be damaged for tables, drawings or images

  • FastPDF: This mode will provide the fastest result for PDF translation by retrieving plain text without taking care of images or formatting-style. This feature works only for Text-Based PDFs (extractable text in PDF). The downloaded translated file will have a .txt extension and will lose all the formatting style of the original text.

  • Auto: the Filter will try to detect if the pdf contains text in this mode, if yes, it will use the Precise PDF mode, if not it will use OCR mode which means the installed OCR filter (IRIS IDRS ©)

When uploading a file in the File Translation menu, users have the choice to select between different modes to translate the PDF file:

../../_images/9.6.Doc9-ug-tools-34.png

Image File Translation

Supported image formats are: bmp, jpg, jpeg, png, tif, tiff

The translated file will be in docx format:

../../_images/Image_file.png

../../_images/Image_file2.png

The following formats are accepted:

  • for bmp: ‘image/bmp’

  • for jpg and jpeg: ‘image/jpeg’

  • for png: ‘image/png’

  • for tif and tiff:’image/tiff’

Please note that he quality of the OCR may vary depending on the image, the text orientation, the fonts used and the language.

Translation with Profile Selectors

In the example below:

../../_images/9Filetranslationselectors.png

‘owner’,’domain’ and ‘size’ are displayed in blue because the user selected them. If no selectors are applied, the system chooses them according to priority rules.

Hover the mouse over the selector to display the label ‘Selected by user’ or ‘Selected by system’:

../../_images/9Selectedbyuser.png

../../_images/Selectedbyuserr.png

Hover the mouse over the information icon to see what profile has been used:

../../_images/9Profileselectionused.png

Translating zip files

To translate zip files:

  1. Click Upload to open the Upload Files dialog

  2. Drag & drop files or click Select Files to select one or more files to translate

  3. Click Upload to upload the files and launch the translation

  4. Click Download to download the zip archive

../../_images/Doc9-ug-tools-66.png

The zip archive (result file) contains all the translated files.

../../_images/Doc9-ug-tools-67.png
5. Click on Edit to see all the extracted files of the zip archive. Files with the status “translated” can be post-edited

Note

  • Unsupported files (unsupported format and/or size limit exceeded) will be ignored, other files will be translated

  • Another archive inside the zip will be ignored

  • If no file exists in the archive or the file format is not supported, the status will be “no file to translate”

  • Corrupted files (or if the archive itself is corrupted), the status will be “error” and no file will be translated

  • If an error occurs during translation:

    • Each file with an error will have the status “error” and won’t be delivered in the zip containing the resulting translations

    • A zip archive can be downloaded if at least one file has been translated

Translating Editor

The Translation Editor is a tool that allows you to review and manage the translated segments of your file. Each segment represents a sentence. You can edit translated segments and come back to them later, you can validate translations and download the translated text with your modifications. Additionally, you can convert a validated translation into a Translation Memory for further use in SES9 or other tools.

Reviewing and managing translations

Shortcuts

In File and Speech Translation editor and Resources editor (UD, Normalization, TM), the shortcuts are shown via the ‘info’ icon:

../../_images/Ses9shortcut.png

../../_images/Ses9shortcut2.png

Display Options

Below the sentences table, you can change the display options:

../../_images/Doc9-ug-tools-16.png

  1. Change the number of sentences displayed per page, so that up to 100 sentences are viewed on the same page

  2. You can also navigate to other pages

  3. You can navigate to following and previous sentences on the same page with the Up and Down arrow keys

Editing Translations

To edit or validate a translation, click on the sentence to translate in the sentence table. The sentence will appear in a text field at the bottom of the page:

  • To modify the machine’s translation, you can type or delete words directly into the text field

  • To validate the machine’s translation, click on the green check-mark next to the Translated Sentence text field. The next sentence to validate will automatically appear in the same Translated Sentence field. Validated edits are saved automatically and you can go back to them at a later date

  • Alternative Translations: If the sentence or a fuzzy match are found in the Profile’s Translation Memory, it will be displayed under the TM tab

../../_images/Doc9-ug-tools-189.png

Use alternative translation button to quickly paste the TM match or Machine Translation in the Target field.

Downloading Translations

Once all translated sentences have been reviewed and validated and you’re ready to transfer the translation from the server to your local system, click Download in the Translation Editor toolbar. The translated file will be regenerated in its original format and the filename will be saved locally as: result_filename.extension

From the Translation Editor tool, you can also create and manage Translation Memory files:

  • Create Translation Memory: Create a TM on the server and assign a name to it

  • Download Translation Memory: Generates a TMX or bitext file to download

../../_images/9Doc8-ug-tools-18.png

Profile selection

Owner selector

Owner corresponds to the owner of the translation resource. It can be SYSTRAN or a non-SYSTRAN owner (self or customer name).

../../_images/9Owner.png

The possible owners include:

  • Self appears only if the user has created a profile and edited the profile options

  • Customer name corresponds to the owner of the SYSTRAN translation resource

  • Systran corresponds to the SYSTRAN Translation Resource

  • Systran and Customer name mean the system will look for a profile created automatically when the customer installs a translation resource and adds at least 1 instance

The following table shows which owner is displayed in the Profile Selection menu by default:

Possible cases

Self

Systran

Default

1

If Self is not available

but Systran is available

then Systran will be displayed

2

If Self is available

but Systran is not available

then Self will be displayed

3

If Self is available

but Systran is available

Neither will be displayed

The priority that applies when translating is as follows:

If Self and Systran are available and neither is selected, then Self is used by default.

Domain selector

Domain is the domain of the corpus used for creating the translation resource. It can be Generic, IT, Tourism etc. For instance, if you wish to translate all languages with a ‘Tourism’ domain profile, simply select ‘Tourism’ in Domain:

../../_images/9Domain.png

The system will automatically look for the corresponding profile available. If no profile exists, another profile will be found according to the priority rules. The following table shows which domain is displayed in the Profile Selection menu by default:

Possible cases

Domain available:

Default Domain display:

1

Generic, IT and Tourism

None

2

Only Generic

Generic

3

Only IT

IT

4

Only Tourism

Tourism

5

etc.

Note

Domain will be empty if more than 1 domain is available..

The priority that applies when translating is as follows:

If several domains are available and none is selected, Generic is used by default. If several Generic profiles exist, the oldest profile (i.e. the profile created first) will be selected.

Size selector

Size describes the size (both in disk space and computing resource usage) of the translation resource. It can be S (for Small) or M (for Medium). Generally S engines use Rule Based Translation engines, while M engines are more likely to be Statistical Engines.

../../_images/9Size.png

The following table shows which size is displayed in the Profile Selection menu by default:

Possible cases

Size available:

Default Size display:

1

S and M

None

2

Only S

Generic

3

Only M

M

The priority that applies when translating is as follows:

If several sizes are available and none is selected, M is used by default.

Note that if a profile S is requested by the user and this profile does not exist, there is no fall back on an M profile.

Translating with Auto-detection

With the profile selection feature, it is possible to define the desired selectors that will apply to all language pairs.

To translate with auto-detection:

  1. Choose Auto-Detect as source language and select a target language

  2. Click on the ‘+’ icon to select the Owner, Domain and Size

../../_images/9Autodetectplus.png

This way, selectors chosen by the user will apply and automatically select the preferred domain profiles when available and fall back to a generic profile for language pairs for which the selected domain is not available.

This feature can be particularly useful when translating a batch of files at once, as the profiles used to translate will be filtered independently of the language pair.