Wednesday, November 24, 2010

Some Light Shed on the “Failed to Leverage Segment” Error in Wordfast Pro

In September I posted a negative review of Wordfast Pro. One of my major complaints was the repeated recurrence of the following error message:

"Failed to leverage segment. The following error occurred: *.txt: Failed to leverage segment. Unknown Wordfast  escape."

The error was particularly frustrating because it occurred on segments that I knew contained an exact or at least a similar match inside my translation memory (TM). In other words, the main reason you use TM technology was being negated.

It turns out that, in my case, the problem’s cause was completely unexpected and, for me at least, difficult to explain. My memories contained thousands and thousands of corrupted lines that were interfering with the adequate functioning of the program. After opening the TM in a word processor, a typical translation unit (TU) turned out to look like this:

20040529=191302      MIGUEL LLORENS  1          ES-EM            "A comienzos de los 90 y hasta el 95 los milicianos protegían el barrio, y ya después robaban y mataban.    EN-US "In the early nineties, up to 1995, the militias protected the neighborhood, and later on they began killing and stealing."   \sectd\linex0\endnhere\sectlinegrid360\sectdefaultcl\sftnbj ; o llamando al (440) 449-9669.  Sherman          - P.V.P:          - P.V.P:

20061127=201125      Miguel Llorens           3          ES-EM            "Es probable que las personas no recuerden exactamente qué hizo o qué dijo usted, pero siempre recordarán cómo los hizo sentir".    &'96; Fuente desconocida  EN-US "People may not remember exactly what you did, or what you said, but they will always remember how you made them feel."    &'96; Source Unknown        \sectd\marglsxn900\margrsxn920\sbknone\linex0\sectdefaultcl\sftnbj ; &t=};usando su SSN o hablando directamente con un operador    ******           Source - P.V.P:
The highlights in yellow are simply gobbledygook pulled from other TUs and metatags. I suspect it was generated during TM conversions from Trados to Wordfast or vice versa. I have heard comments that the .tmx standard, which theoretically should make memories easy to transport from one program to another, is still more of an aspiration than a reality. Perhaps this is a symptom of that.

In any case, the problem is relatively easy to fix. Simply open the TM file with Word. Select the gibberish in one of the units, and copy and paste it into Replace. You have to include the tabs and any random formatting garbage in order to make the cleansing complete. Leave the “Replace with” window blank and run the function. This will wipe the gibberish from your memory.

In my case, with my mother-of-all-translation-memories, each replacement wiped tens of thousands of instances of the corrupt lines from the memory. However, in my case, there were many different types of corrupt lines. So I had to scan the file manually and spot the bad lines. The cleanup took several hours from my life which I’ll never get back, but the program works about 80% better now.

In general, I still have some quibbles about the program's speed and the frequency with which Wordfast freezes up. But at least it is not the horror show reviewed several months ago. While I still wouldn’t recommend it, I am now looking forward to future updates to see whether it matches its previous degree of usability.

No comments: