The ATA Chronicle - November/December 2023 - 29

RESOURCE REVIEW By Richard Sikes
Adventures in the World of TMX
Let's discuss some of the issues encountered when converting legacy TMX
files from some competitive products to a flavor of TMX for input into memoQ
TMS. We'll explore some of the differences between how several TMS
technology manufacturers have interpreted the TMX standard. The conversion
direction covered here is from the legacy formats to memoQ-friendly TMX,
but the same principles apply regardless of the conversion direction.
his article is not
about memoQ, per se.
Nor, with one small
exception, is it intended
to criticize or indicate
a preference for one
T
interpretation of the standard
over another. The legacy
formats that were converted
in the examples used here
were generated from XMLbased
LogiTerm bitexts and
TMX output produced by
XTM, Memsource (Phrase),
and Trados.
The goal in any translation
memory conversion is to
achieve a state in the new
TMS that yields 100%
leverage when compared
with the old tool. It's more
of an approximation, not
an absolute 100%, but we
should strive for the best
result possible, at least with
a representative sampling of
legacy files.
There are three primary
components to conversion:
y The formal structure
of the TMX export
vis-à-vis the structure
needed for import.
y Tag representation
and handling.
y Segmentation differences.
We'll concentrate here
primarily on the structure of
TMX files. The notion of tag
conversion is complex and
www.ata-chronicle.online
American Translators Association 29
mostly beyond the scope of
this article, but we'll touch
on it. Because segmentation
can be defined on a projectby-project
basis in some
TMSes, variations can be
both between and within
translation memories that
come from the same tool.
Generally, we don't address
segmentation, but we need to
be aware of the consequences
in terms of potential matching
loss. This can, of course,
remain true even when using
translation memories in the
same TMS on a different
project or document that is
segmented differently.
memoQ TMX
The TMX standard is
somewhat loosely defined.
TMS providers have
interpreted the standard
differently from one another.
As a place to start, we can
look at what memoQ TMX
requires from the standpoint
of structure. This is the
structure we need to match as
closely as possible to achieve
optimal leverage when
converting a TMX file created
by a different TMS.
A TMX file is XMLbased,
so there is an XML
declaration at the top of
the file. This is common
to all TMX files. After the
declaration, memoQ TMX
files contain two main parts:
the header and the body
sections. (See Figure 1.)
The header node contains
metadata about the TMX
file itself. Of the several
attributes in the header node,
one of the most important
is the " srclang, " which
specifies the language in
which the source content was
written. The other attributes
in the header node are not
important for conversion.
The header node contains
child nodes that, according to
the TMX standard, can only
be  (property) nodes.
Generally in XML, properties
are used for programmatic
control of such things as
color, visibility, position,
numeric display format, etc.
In TMX, these nodes in the
header are used to retain
metadata pertaining to the
provenance of the content.
The  nodes are defined
by type attributes. The name
and meaning of the type
attributes-hence the usage
of the nodes-are decided by
the TMS provider.
In memoQ TMX, the child
nodes represent the four
built-in metadata fields
that memoQ TMS uses
throughout for automated
matching of translation
memories, termbases, and
projects: client, project,
domain, and subject.
There is also a description
field. These five fields are
automatically populated
when the equivalent fields are
populated in the project for
which a translation memory
is originally created. There is
also a target language node
and the translation memory
name node.
The body section follows
the header section. (See
Figure 2 on page 30.) The
body contains the translation
units. Each translation unit
contains a metadata section
plus one section each for the
source and target segments.
Each translation unit
FIGURE 1 Example of the header node in a memoQ TMX file
http://www.ata-chronicle.online
The ATA Chronicle - November/December 2023

Table of Contents for the Digital Edition of The ATA Chronicle - November/December 2023
