Translation Memory Without Alignment

T o b i a s wrote:

I suspect many of your problems are the result of messy/noisy PDF files/conversions and the resulting huge files.

Studio's built-in PDF converter, while very good (I purchased the stand-alone version to get better performance), tends to insert tags for the most trivial reasons (word-spacing is a particular annoyance) The resulting Trados file may easily contain 1000s more tags than it otherwise would. For this reason, with a 300 page file, you might want to consider a separate tool to do the conversion, e.g. ABBYY Finereader. With a 300 page file, at a minimum, you will want to clean it of superfluous tags to make it more easy to handle. Consider a tool like Code Zapper, which will jettison unneeded tags, and will also export/re-import the heavy graphics temporarily while you translate. Also consider breaking up the Word file into different pieces. As far as alignment goes, the new aligner in Studio 2014 is reportedly very good. LF_Aligner is also an excellent product.

Andrew052 wrote:

I would be grateful for your comments on this topic.

I have received 10 pdf files (both originals and translations). The originals are in pdf and the translations are in word. The files contain a lot of graphics (brochures, etc.). The translations were done without using any cats, when a pdf is converted into word. I need to identify any matches across these files and the translated files.

I basically need to find two solutions:

1) Capability to concordance thru the files

2) Capability to identify any matches of higher than 70

Solution that I have used before:

1.

1) (Trados 11) I create a TM with the following settings: (in the tab "Fields and Settings") Name: someting like "Number in the Register"; Type: Number.

2. I create a list of folders with the numbers and place the original and translation for each file in separate folders.

3. I create a separate project for file No. 1 in Trados. Settings: (Pre-Translate) Copy source to target if no match has been found. I open the file and click on the number of the first segment, then I scroll down and click CTRL+click to select all segments. Then I change the status of all segments to "translated".

4. I update the main memory (via the batch task "update main memory"). Settings: (Tab "Translation Memories and Automated Translation", sub-tab "Update") this TM (which I have created in point 3) enables to type the number "1" under the word "Value" in this window.

When I open a file that has some matches with file No. 1, the matches are shown in the matches window in the following way: original text on both side and the text (Number in the Register: 1) to the right.

Thus, I know that I have to open file no 1 (original or translation) and copy-paste the match.

Benefits:

I have done this procedure for 5 projects, when the customer had no TM in place. I had to handle files of less than 50 pages each. It enabled me to quickly identify any matches and incorporate them. It took me a couple of hours to create these TMs, while I was doing other work on the PC -- I was just clicking a couple of times every 10 minutes. Thus, it means a couple of clicks per file vs. aligning the text in a file by hand.

Problems:

Now I have to handle files, each of which has 300 pages. When I try to just open a file in Trados, it takes 2 hours just to open the file (I have 4 G RAM, so it makes the process longer).



How do you tackle this issue? (Such options as aligning by hand or via other means, copying the materials to the same file and marking particular parts of the file (using colors or other means) are not feasible due to time constraints.