TM performance tuning
You can tune the performance of a translation memory (TM) in terms of: performance, fuzzy index and statistics.
SDL Trados GroupShare performs the re-indexing and recomputation tasks in the background, leaving you free to do other actions. Monitor their progress in the System Configuration view by going to the Background Tasks tab.
Performance
Trade off speed versus accuracy by changing the minimum match score.
- When the match score decreases, the fuzzy search optimizer narrows its search.
- When the match score increases, the fuzzy match search is broadened.
Fuzzy index statistics
Words that occur very frequently in translation units (TUs) are not useful for fuzzy indexes. Typically, these are words like a, the, to. Recompute fuzzy index statistics removes the 1000 most frequent words from the fuzzy indexes.
This can make a difference in performance for large TMs. If the TM has less than 50,000 TUs, it will not make a noticeable difference.
Re-index TM
TM are re-indexed from scratch. The re-indexing process goes through each TU, finds the tokens (words), and then computes the fuzzy indexes.
Do this, for example, if you enable recognition of dates or variables after you change token recognition.
This option includes the previous option (Recompute fuzzy index statistics).
On large TMs, re-indexing can take a long time.
upLIFT alignment status
upLIFT alignment is the fragment alignment (also known as FGA) that automatically occurs for TMs which have 5,000 TUs or more. The automatic alignment happens between 00:00 AM and 06:00 AM. You can check whether the TUs are aligned or not on the Performance Tuning page of server-based TM s which meet the minimum 5,000 TU requirement.