1. Do you guarantee the quality of the translation memories?

2. How does the quality of these aligned translation memories differ from the quality of the licensed translation memories?

3. What is the legal situation? Isn't the content of the translation memories copyrighted?

4. Couldn't I create those translation memories myself?

5. What kind of benefit am I likely to receive from an external translation memory?

6. Are there any other uses aside from the traditional perfect and fuzzy matches with my translation memory tool?

7. Can I merge the purchased translation memories with my existing translation memories?

8. How much does purchasing a translation memory cost?

9. Can you help me to integrate the translation memories into my workflow, or can you help me to set up a workflow from scratch?

1. Do you guarantee the quality of the translation memories?

We look at three levels of quality:

1. Structural integrity
2. Linguistic quality
3. Matching ratio

We guarantee the structural integrity of the data. This means that the TMX file you receive will have no file-internal corruptions and the file will adhere to the TMX standard (see www.lisa.org/standards/tmx). The code page of the file will be Unicode (in non-Asian languages it will be UTF-8, and in Asian languages UTF-16).

Since we use only already-published data as the basis for our translation memories, the linguistic quality is as good as the published data. We don't perform any linguistic processes on matching data.

The matching ratio—the percentage of matched-up source and target sentences—is very high. We can achieve this because we use a combination of sophisticated tools and workflows with which we are able to align numerous languages simultaneously.

To verify the accuracy, we have hired independent linguists to check the match ratios in language combinations where at least one language was not known to the person performing the alignment. Their results show an  accuracy ratio of 99.4%.

back to top

2. How does the quality of these aligned translation memories differ from the quality of the licensed translation memories?

The only point where the quality of the aligned translation memories differs from the licensed memories is in the area of matching ratio (see above).

back to top

3. What is the legal situation? Isn't the content of the translation memories copyrighted?

We have put together a White Paper that explores the legal situation in detail.

To summarize the legal position, the documents that form the basis of the translation memories may be copyrighted, but the TMs themselves are not. The TMs are not copyrighted because it is not possible to use them to recreate the original texts. We make sure of this by scrambling the order of the segments and by limiting the lengths of the segments we use.

back to top

4. Couldn't I create those translation memories myself?

Yes, you could. However, most anyone who has performed alignment will tell you what a tedious and time-consuming task it is, especially if you don't have the right tools and processes in place. Purchasing a translation memory from us will prove to be significantly more cost-effective. (Feel free to prove us wrong on this point.)

back to top

5. What kind of benefit am I likely to receive from an external translation memory?

First of all, because the translation memories are industry-specifically categorized, there will be significant overlap between the subject area of your translation and that of the external translation memory. While it is unlikely that you will have large passages of your translation pre-translated with perfect matches, you should view the translation memory primarily as a linguistic resource that will allow you to unify your terminology with that of the owner of the database.

All major translation memory tools offer a feature that is generally known as a "concordance search," the ability to look for a particular term or phrase within the context of a translation segment. This is one way to quickly access the linguistic wealth of an external database.

Another more sophisticated manner of mining terminology from existing translation memories is with the help of term extraction tools, such as SDL MultiTerm Extract®, SDLPhraseFinder®, or MultiTrans®. These tools allow you to quickly assemble glossaries from existing bi- or multilingual translation memories.

back to top

6. Are there any other uses aside from the traditional perfect and fuzzy matches with my translation memory tool?

Yes! Aside from the term extraction purposes described above, there are a number of other uses of translation memories.

TM-based authoring is an exciting new technology that essentially allows you to use the same technology used by the translator in employing translation memories in the authoring process. As a result, you will not only achieve greater consistency within your source texts, but because your authors have used the same translation memory for creating the texts as your translators will use to translate them, you will achieve extremely high repetition rates. Commercial tools that support this process include Authoring Coach® or SDL AuthorAssistant®.

The latest generation of machine translation applications uses bilingual data to train its recognition in specific areas. The best-known commercially available application in this area is Language Weaver®.

back to top

7. Can I merge the purchased translation memories with my existing translation memories?

Yes, you can. Unlike the licensed translation memories, there are no limits to what you can do with this data as long as you don't redistribute or resell it.

back to top

8. How much does purchasing a translation memory cost?

You can find pricing information for existing translation memories here.

back to top

9. Can you help me to integrate the translation memories into my workflow, or can you help me to set up a workflow from scratch?

Yes, we would be glad to help you with that. We offer either remote or on-site assistance. For more information, contact service@tmmarketplace.com.

back to top