Cracking the Metadata Tower of Babel

Typically, metadata (i.e., machine-readable descriptions of digital resources) include natural language fields. Titles, summaries, or free-text keywords are common examples of such metadata elements.

In a multilingual context, correctly processing these free-text elements requires us to accurately identify the language of each text.  In metadata standards such as IEEE LOM, this is achieved by using a special construct called “language strings” where the text language is systematically identified.

An example of how language strings can be implemented in XML is shown below:

<metadata>
    <title>
        <string language="en">Title</string>
        <string language="fr">Titre</string>
    </title>
    <description>
        <string language="en">Description</string>
        <string language="fr">Description</string>
    </description>
</metadata>

This approach is not ideal because it requires one to parse the full metadata record to obtain a complete description in a given language.

This is why we usually recommend the use of language blocks where all the metadata elements in a given language are grouped together, which makes them easier to access and process.

<metadata>
    <block language="en">
        <title>Title</title>
        <description>Description</description>
    </block>
    <block language="fr">
        <title>Titre</title>
        <description>Description</description>
    </block>
</metadata>

The example above gives a sense of how language blocks can simplify processing.  In this way, displaying a complete description in a given language requires only the extraction of a corresponding language block.

In the same way, adding a language translation can be easily achieved by submitting a block in a source language to a translation service and adding back the translated block to the record.