Unbabel supports the translation of a large number of file formats. Different formats have different structures, rules and properties, which created different requirements when it comes to select which content needs to be translated and which one does not. This is done through Filter configurations. These configurations also helps the translation pipeline to deliver a file that respects and resembles the format of the original source.
These configurations will behave differently according to the file type being translated and its complexity. While simpler files such as plain text (txt) will mostly see the full extent of the text content extracted (i.e., selected for translation), translated and delivered, formats such as docx or or xlsx have more complex rules. Understanding these rules guarantees you can translate a file that is properly formatted and that you are in full control of the content that is being changed during translation. The configurations properties also enable you to manipulate the source file in ways you can find convenient - such as preventing specific content from being translated (protected content).
Below we list all the supported file formats and the properties of their standard filter configurations. When more information is available, click on the file format name to see additional information.
Supported file formats
CSV (Comma Separated Values)
Note: More information about the CSV filter is available in this article.
The filter extracts all table data from all columns. Generic placeholders are protected, as well as embedded HTML.
Technical information:
- field delimiter is the comma ‘,’
- text qualifier is the double quote “”
- csv escaping mode - duplicates qualifier
- excludes qualifiers from extracted text
- excludes leading/trailing white spaces from the extracted text
- adds qualifiers to output when appropriate
- extraction mode - extracts table data
- table properties - values start at line 1 (no column with names)
- extracts data from all columns
- the number of columns is defined by values (may vary in different rows)
- allows trimming of leading/trailing spaces and tabs
- converts \t, \n , \\ and \uXXXX into characters
- separates lines with line-feeds (\n)
- includes okf_html@FP-subfilter-default
- protects generic placeholders
DITA (Darwin Information Typing Architecture)
The filter accepts only well formed xml documents (which adhere to specific dita syntax rules). Generic placeholders are protected.
Technical information:
- assumes the document is well formed
- preserves white space
- uses codeFinder to protect generic placeholders
DITAMAP (Darwin Information Typing Architecture Map)
The filter accepts only well formed documents (which adhere to specific syntax rules).
Technical information:
- assumes the document is well formed
- lists elements and attributes for translation
DOCM (Microsoft Word)
The filter extracts everything except document properties, comments, graphical metadata. It automatically accepts revisions if they are present in the document.
Technical information:
- does not extract document properties and comments
- translates headers and footers
- excludes graphical metadata
- automatically accepts revisions and extracts their content
- includes styles and highlights
DOCX (Microsoft Word)
The filter extracts everything except document properties, comments and graphical metadata.
- extracts headers and footers
- excludes graphical metadata
- includes HTML subfilter
DTD (Document Type Definition XML)
The filter is intended to process XML-DTD that have translatable text entity declarations.
HTML/HTM (HyperText Markup Language)
The filter extracts all content from the file but tags are not translated. Generic placeholders are protected.
Technical information:
- protects generic placeholders
ICML (InCopy Markup Language)
The filter extracts all content from the file.
Technical information:
- extracts master spreads
- simplifies inline codes where possible
- uses codeFinder for tag protection
IDML (InDesign Markup Language)
The filter extracts all content from the file, except for XML structures.
Technical information:
- does not untag XML structures (the filter cannot put the tags back. This needs to be done in a manual DTP process, which depending on the size of the file might be an issue)
- extracts master spreads
JSON (JavaScript Object Notation)
The filter extracts all values. Embedded HTML and generic placeholders are protected.
Technical information:
- extracts all key/string pairs
- extracts strings without associated key
- uses key as resname
- has html subfilter which deals with embedded html and protects generic placeholders
Markdown/MD (Markdown)
The filter extracts all content from the file. Embedded html and generic placeholders are protected, except placeholders of type #company and [checkout_date], as # and [...] are part of markdown syntax.
Technical information:
- translates fenced code-blocks
- translates inline code blocks
- translates YAML metadata header
- translates image alt text
- placeholders are protected as inline codes. For this configuration placeholders of type #company and [checkout_date] are not protected as # and [...] are part of markdown syntax.
- uses the default embedded HTML filter configuration tailored for the Markdown filter (no html subfilter is needed)
MIF (Adobe FrameMaker Interchange format)
The filter extracts variables, index markers, body pages and master pages.
Technical information:
- extracts variables
- extracts index markers
- extracts body pages
- extracts master pages
- inline code protection for fonts
MQXLIFF (XML Localization Interchange File Format)
The filter extracts all content from the file. Generic placeholders are protected.
Technical information:
- adds the target language attribute if not present
- segments only if the input text is segmented
- includes ITS markup
- balances codes
- uses a custom xml stream parser
- protects generic placeholders
MXLIFF (XML Localization Interchange File Format)
The filter extracts all content from the file. Generic placeholders are protected.
Technical information:
- adds the target language attribute if not present
- segments only if the input text is segmented
- includes ITS markup
- balances codes
- uses a custom xml stream parser
- sets Finished segments as translate=”no”
- protects generic placeholders
ODP (OpenDocument (Ver 2) Presentation)
The filter extracts everything from the file. All the different embedded files are treated as sub-documents by the filter. This means that, for example, when represented in XLIFF, a single ODT extracted to a single XLIFF document is made up three XLIFF <file> elements: One for content.xml, one for style.xml, and one for meta.xml. Note that very often, only content.xml has extracted text.
Technical information:
- extracts everything
ODS (OpenDocument (Ver 2) Spreadsheet)
The filter extracts everything from the file. All the different embedded files are treated as sub-documents by the filter. This means that, for example, when represented in XLIFF, a single ODT extracted to a single XLIFF document is made up three XLIFF <file> elements: One for content.xml, one for style.xml, and one for meta.xml. Note that very often, only content.xml has extracted text.
Technical information:
- extracts everything
ODT (OpenDocument (Ver 2) Text Document)
The filter extracts everything from the file. All the different embedded files are treated as sub-documents by the filter. This means that, for example, when represented in XLIFF, a single ODT extracted to a single XLIFF document is made up three XLIFF <file> elements: One for content.xml, one for style.xml, and one for meta.xml. Note that very often, only content.xml has extracted text.
Technical information:
- extracts everything
OTS (OpenDocument (Ver 2) Spreadsheet)
The filter extracts everything from the file. All the different embedded files are treated as sub-documents by the filter. This means that, for example, when represented in XLIFF, a single ODT extracted to a single XLIFF document is made up three XLIFF <file> elements: One for content.xml, one for style.xml, and one for meta.xml. Note that very often, only content.xml has extracted text.
Technical information:
- extracts everything
PO (Portable Object)
The filter treats the file as bilingual - it extracts the content of "msgid" and places the translation in "msgstr". Generic placeholders are protected.
Technical information:
- does not extract document properties and comments
- extracts Masters
- ignores placeholder text in Masters
POTM (Microsoft PowerPoint)
The filter extracts all content but document properties, comments and notes. It extracts the content of the Master slide while ignoring the placeholder text in it.
Technical information:
- does not extract document properties and comments
- extracts Masters
- ignores placeholder text in Masters
POTX (Microsoft PowerPoint)
The filter extracts all content but document properties, comments and notes. It extracts the content of the Master slide while ignoring the placeholder text in it.
Technical information:
- does not extract document properties and comments
- extracts Masters
- ignores placeholder text in Masters
PPSM (Microsoft PowerPoint)
The filter extracts all content but document properties, comments and notes. It extracts the content of the Master slide while ignoring the placeholder text in it.
Technical information:
- does not extract document properties and comments
- extracts Masters
- ignores placeholder text in Masters
PPSX (Microsoft PowerPoint)
The filter extracts all content but document properties, comments and notes. It extracts the content of the Master slide while ignoring the placeholder text in it.
Technical information:
- does not extract document properties and comments
- extracts Masters
- ignores placeholder text in Masters
PPTM (Microsoft PowerPoint)
The filter extracts all content but document properties, comments and notes. It extracts the content of the Master slide while ignoring the placeholder text in it.
Technical information:
- does not extract document properties and comments
- extracts Masters
- ignores placeholder text in Masters
PPTX (Microsoft PowerPoint)
The filter extracts all content but document properties, comments and notes. It extracts the content of the Master slide while ignoring the placeholder text in it.
Technical information:
- does not extract document properties and comments
- extracts Masters
- ignores placeholder text in Masters
PROPERTIES
The filter extracts the content of the values. Embedded HTML and generic placeholders are protected.
Technical information:
- uses localization directives when they are present
- extracts items outside of the scope of localization directives
- extracts comments to note properties
- converts \n and \t to line break and tab
- CodeFinder takes care of placeholders (an html subfilter deals with the embedded html)
- does not escape extended characters (\uHHHH notation)
RESX (.NET Managed Resource)
The filter extracts the content of the values. Embedded HTML and generic placeholders are protected.
Technical information:
- extracts by default //data[not(@type) and not(starts-with(@name, '>'))]/value and //data[@name='$this.Text']/value
- extracts as notes //data[not(@type) and not(starts-with(@name, '>') or starts-with(@name, '$'))]/value
- an html subfilter deals with placeholders and embedded HTML
SDLXLIFF (SDL XML-based Localization Interchange File Format)
The filter extracts all content from the file. Generic placeholders are protected.
Technical information:
- uses SDLXLIFF writer
- adds the target-language attribute if not present
- preserves whitespace by default
- skips seg-sources with no marked segments
- segments only if the input text unit is segmented
- includes ITS markup
- balances codes
- uses a custom xml stream parser
- sets Finished segments as translate=”no”
- protects generic placeholders
SRT (SubRip Subtitle)
The filter extracts all content from the file. Generic placeholders and line-breaks are protected.
Technical information:
- a regex filter processes the .srt whilst the html subfilter deals with embedded html and protects generic placeholders
- the time-codes are not added as notes due to a limitation we found when using regex filter + html subfilter
STRINGS (Text Strings File)
The filter extracts the content of all values. Embedded HTML, generic placeholders and line-breaks are protected.
Technical information:
- does not include notes (limitation we faced when using regex filter + html subfilter)
- extracts the content of the source group
- preserves whitespace
- regular expressions options: dot also matches line-feed + multiline
- uses localization directives when they are present
- extracts items outside of the scope of localization directives
- escaped characters use backslash
- mime type for the document: text/plain
- protects generic placeholders and embedded HTML
STRINGSDICT (Apple Stringsdict)
The filter extracts the content of <string> elements (without elements with keys NSStringFormatSpecTypeKey and NSStringFormatValueTypeKey). Generic placeholders are protected.
Technical information:
- extracts for translation /plist/dict/dict/string and /plist/dict/dict/dict/string
- does not extract strings with keys NSStringFormatSpecTypeKey and NSStringFormatValueTypeKey
- protects generic placeholders
TMX (Translation Memory eXchange files)
The filter extracts all content from the file.
Technical information:
- groups all document parts skeleton into one
- skips invalid TUs
- creates the segment if segtype is ‘sentence’ or is undefined
- string used to delimit property values when there are duplicate properties: ','
TXT (Plain Text)
The filter extracts all content from the file. Generic placeholders are protected.
Technical information:
- extracts text by lines
- converts \t, \n, \\ and \uXXXX into characters
- separates lines with line-feeds (\n)
- protects generic placeholders
VSDX (Microsoft Visio)
Technical information:
- uses the default okp_openxml filter
- includes HTML subfilter
- it offers no specific options for Visio
XLF/XLIFF (XML Localization Interchange File Format)
The filter extracts all content from the file. Generic placeholders are protected.
Technical information:
- adds the target language attribute if not present
- segments only if the input text is segmented
- includes ITS markup
- balances codes
- uses a custom xml stream parser
- sets Finished segments as translate=”no”
- protects generic placeholders
XLSM (Microsoft Excel Macro-Enabled)
Technical information:
- does not extract document properties and comments
- does not extract hidden rows or columns
- does not extract sheet names
- does not extract diagram data
- does not extract drawings
- The html sub-filter deals with embedded html and protects generic placeholders.
XLSX (Microsoft Excel)
The filter extracts all content from the file but document properties, comments, hidden rows/columns, diagram data and drawings. Embedded HTML and generic placeholders are protected.
Warning: XLS, the old format used by pre-2007 MS Office, is not supported. Please convert any of these files to XLSX
Technical information:
- does not extract hidden rows or columns
- does not extract sheet names
- does not extract diagram data
- does not extract drawings
- embedded HTML and generic placeholders are protected
XLTX (Microsoft Excel)
The filter extracts all content from the file but document properties, comments, hidden rows/columns, diagram data and drawings. Embedded HTML and generic placeholders are protected.
Technical information:
- does not extract hidden rows nor columns
- does not extract sheet names
- does not extract diagram data
- does not extract drawings
- includes HTML subfilter
XML (Extensible Markup Language)
The filter accepts only well formed documents (which adhere to specific syntax rules). HTML is protected only in CDATA . Generic placeholders are not protected.
Technical information:
- accepts only valid, well-formed XML
-
protects html only in CDATA
-
does not protect placeholders
- preserves whitespace
YAML/YML (YAML Ain't Markup Language)
The filter extracts the content of all values in the file. Embedded HTML and generic placeholders are protected.
Technical information:
- extracts isolated strings
- extracts all pairs
- uses key as name
- uses the full key path
- does not use codeFinder
- The html subfilter deals with placeholders and embedded html.
VTT (Web Video Text Tracks)
The filter extracts all content from the file. Embedded HTML, generic placeholders and linebreaks are protected.
Technical information:
- extracts the content of the source group using regex
- preserves whitespace
- regex options: dot also matches line-feed + multi-line
- uses localization directives when they are present
- extracts items out of the scope of localization directives
- beginning/end of string: “”
- escaped characters use backslash prefix
- mime type: text/plain
- The html subfilter deals with placeholders and embedded html
TSV (Tab Separated Values)
The filter extracts all table data from all columns. Generic placeholders are protected.
Technical information:
- field delimiter - tab ‘\t’
- extraction mode - extracts table data
- table properties - values start at line 1 (no column with names)
- extracts data from all columns
- the number of columns is defined by values (may vary in different rows)
- allows trimming of leading/trailing spaces and tabs
- converts \t, \n , \\ and \uXXXX into characters
- separates lines with line-feeds (\n)
- protects generic placeholders
What are generic placeholders?
Generic placeholders are character combinations that will ensure that the text in your document is not picked for translation, and instead is delivered as in the original source, placed in a correct and syntactically correct position in the translation. When a filter is able to handle these placeholders, we will mention it in the description above. If there are exceptions, they will be stated in the technical information.
The list of placeholders to be used with the Standard Filter configuration is:
| Placeholder pattern | Example |
| {placeholder} |
Dear {first_name}, thank you for contacting. |
| ${placeholder} | This will be sent to ${package_destination}. |
| $((placeholder)) | Your voucher code is $((bonus_code)). |
| {{placeholder}} |
Esteemed {{contact.name}}, |
| %#@placeholder@ | Please submit it by %#@end_date@. |
| @PLACEHOLDER | Your shipment @PACKAGE_ID arrived. (requires capitalization) |
| #placeholder |
Please reach out to #department_name. (does not work with diacritics) |
| %%placeholder%% | Have your %%product.name%% today! |
| %placeholder | Click on %site.element to %action.1 |
Standard vs Custom filters and File Engineering Services
Standard filters are available for any Unbabel customer that has just started to translate Projects. However, some files might have requirements that are not covered by the standard configuration. When this is the case, we can either manipulate the file outside the standard configuration - usually if it's a one-time event translation - by purchasing a File Engineering service or create a custom filter that specifically accommodates the file requirements for the current and future translations.
Warning: Custom Filters and File Engineering may not be available for your subscription tier. Speak to your Account Manager/representative to know whether you can benefit from them.
Filter configurations outside the standard one will be looking for a specific file structure - as they were tailor made to fit certain file layouts or content distribution. It is therefore important that you make sure the file you are translating is the correct fit for selected configuration.
Warning: If you are submitting a file that is different from your usual content for the first time, please check the configuration description to ensure it will translate correctly.