Hi! I've got a sentence starting with "A variable speed drive".
I use a fragment TM for term recognition, and "variable speed drive" is included as a term in that TM. Now, instead of recognizing "variable speed drive" it suggests a prefix match based on "A variety", which is another term in the Fragment TM.
My settings for the fragment TM are as follows:
Memory type: Fragments memory
Matching type: fuzzy
Prefix matching checked, 40%
min. prefix length: 4
Greedy exact matches checked.
What settings should I use to make CT recognize "variable speed drive" in addition to any other (prefix) matches? I tried switching greedy matches off, but it doesn't make any difference.
Thank you for your help!
Hello Stefan! It is a bit hard to tell without seeing the detailed list of the prefix matches for the segment, in the tab for this TM (not just the Matchboard). Maybe you are just looking for the match in the Matchboard only? Also, I would recommend using the "Fixed length" prefix option instead of the percentage-based. It is much more accurate. You need to reload/restart to change the prefix option.
Hello Igor, many thanks for your reply. I played around some more with the TM settings and found a combination of settings that works for me - I just increased the min. prefix length to 5. I left the percentage-based setting for now but will also try the fixed length option.
Hi, it's me again with a problem thst is somehow similar to the one mentioned above:
Sometimes, terms that exist in my fragment TM overlap in a given source sentence, such as:
"Recognize a structure, or more generically a certain property of a surface"
In my fragment TM, "or more" and "more generically" are individual terms, and in my example only "or more" is recognized, not "more generically".
I have encountered this several times that when two terms overlap in this way, only the one that comes first in the sentence is recognized, not the second. Shouldn't this be solved by switching "greedy exact matches" off? Switching it on or off doesn't make any difference in my case.
I would be grateful for any help. Many thanks in advance!
The greedy exact matching option handles the situation when you have full exact TM match for a given sentence, and in addition, you wish to check if there are any useful fragment matches inside this exact match.
The fragments overlapping is tricky to solve in the context of auto-assembling, matching efficiency/speed. If CTE was able to recognize fragments embedded within the larger fragments or both sides of an overlapped larger fragment, the program would be slower, the auto-assembling algorithm would be in trouble to pick the right fragment (e.g. the left or right one). In other words, the amount of possible fragments to search for and display would be really huge.
You can solve your issue in two ways. Either add the longest possible fragment (e.g "or more generically") to your TM/glossary to cancel the overlapping for this fragment, or put the unrecognised part of the segment into another glossary.
Thanks Igor, I will just add the longest possible fragement, as you suggested.