Note to translators: speeding up ABBYY Lingvo and other tools
Electronic dictionary today has become one of the main tools of the translator, along with the browser, digitized reference books and correspondence databases (the latter is important mainly for translators of non-artistic texts).
Among the electronic dictionaries, ABBYY Lingvo is distinguished by one key feature: full-text search with indexing. Something similar can be implemented with the help of indexes in Adobe Acrobat, but the convenience of the interfaces in the dictionary field is beyond comparison.
ABBYY Lingvo has long been transformed from a regular dictionary into a universal source aggregator. In addition to the creator’s titanic work, enthusiasts digitized hundreds of manuals in Lingvo format, including basic bilingual dictionaries, and multi-volume dictionaries of the Cambridge, Collins, Longman, Merriam-Webster, Oxford, and encyclopaedic dictionaries like Britannica. Local copies of network giants have been created (Wikipedia, Wiktionaries, Urban Dictionary, and so on). And in normal use, this would provide exceptional opportunities. But with full-text search, all this wealth turns into language corpuses and correspondence bases. The value of such a search when translating complex terms, stable phrases, phraseological units is difficult to overestimate.
')
With each release, ABBYY expands the allowable limits of compiled dictionaries and search indexes. Already, you can compile a dictionary of almost 2 gigabytes of source text. However, when connecting a large number of dictionaries, the index grows. Both the dictionaries on the disk and the user search index can also reach gigabyte sizes. In this case, full-text search slows down, the speed of work of the hard drives starts to influence it. The era of SSD can help in solving this problem, but so far these mechanisms are not used everywhere because of the higher price and less wear resistance. Fortunately, there is a way that wins even speed SSDs.
You can force the program to move the dictionaries and search indexes themselves from disk to memory. Rather, we ourselves will move them there, and the program will continue to work with them as if they are still on the hard drive. The old proven tool will help us - a RAM disk.
There are many programs that create the similarity of a disk in memory. Special attention should be paid to those that allow the use of memory, inaccessible to other needs.
In modern 32-bit non-server Windows, there are two main limits of memory usage: software restriction to four gigabytes , which cuts down to three memory allocation features for devices . For these reasons, no matter how much memory you put, the main Windows user will not see more than 3 gigabytes plus / minus a little more.
However, these limitations have their own workarounds: PAE and memory remapping (memory hole, PCI hole ). Therefore, if you have a non-64-bit or non-server system and at the same time three gigabytes of memory is not enough to accommodate a RAM disk with everything you need, you will have to turn on all these things and increase the memory. It is better to check the BIOS yourself: for example, in this useful article certain devices are denied support for the described workarounds, however I found the “memory hole” option in the bios of my old doomed motherboard, after switching on which everything worked as it should.
Before all the described manipulations, it is useful to archive important data and create an up-to-date image of the system for recovery if something goes wrong.
To date, there are two well-known programs that allow you to use all the invisible memory for the system: Primo Ramdisk (VSuite Ramdisk II) and SuperSpeed ​​RamDisk Plus . If anyone plans to do with three gigabytes, you can pay attention to the very fast (and according to some data - see here and here - the fastest) RAMDisk Enterprise . All these programs allow you to dump the contents of RAM disks into images on the hard drive before shutting down the computer or restarting and restoring the images to memory at the earliest stage of the system boot. Although this will slow down the download itself, the future gain is worth even a minute delay at launch.
Most motherboards for desktops and laptops of the last 6-7 years allow you to install up to 8 gigabytes of memory. Now it is a fairly inexpensive pleasure, and this size should be enough even for the richest set of dictionaries in Lingvo. However, in extreme cases, you can get by with a smaller size.
What can be moved to Lingvo on a RAM disk?
1. The simplest is dictionaries, both user and system. The translator connects user dictionaries by itself, they can be located anywhere. System dictionaries on Windows XP are located in the folder c: \ Documents and Settings \ All Users \ Application Data \ ABBYY \ Lingvo \ version number \ Dic \ System.
Both groups of dictionaries can be left in their original place for backup, and the program can point to a copy folder using dictionary configuration files. For user dictionaries, this file is% USERPROFILE% \ Local Settings \ Application Data \ ABBYY \ Lingvo \ version_name \ Dic \ dictconf.ini, for system dictionaries - c: \ Documents and Settings \ All Users \ Application Data \ ABBYY \ Lingvo \ version_DiCy \ Dic \ dictconf.ini. In both files, it is necessary to replace the old addresses with new ones with the help of autochange, before not forgetting to save backup copies of both ini-files. If this method seems inconvenient, you can resort to the following.
Naturally, all the manipulations proposed in this and the following paragraphs should be done after closing the program and unloading it from memory (Lingvo can complete work for a long time, check better for programs that display the activity of processes).
2. You can move the folder with the program and binary libraries (c: \ Program Files \ ABBYY Lingvo version_number). Perhaps this will not give an impressive increase in speed, but something can be won. Moreover, this folder also contains numerous .amd and .amm files responsible for morphology. If the program itself does not load them into memory completely, but repeatedly loads from disk if necessary, then the movement makes sense.
3. And finally - vocabulary indexes. For system dictionaries, this will be the c: \ Documents and Settings \ All Users \ Application Data \ ABBYY \ Lingvo \ version_name \ Index folder, and the user’s directory is% USERPROFILE% \ Local Settings \ Application Data \ ABBYY \ Lingvo \ version_Number \ NonAbbyyIndex (although move the entire parent folder; for system dictionaries it is better not to do this, because along with the index there are sounds in the parent folder that weigh a lot and are rarely needed).
There should be another difficulty. If moving dictionaries can be supported by editing configuration files, then moving indexes and the program itself is more intricate. But here we will be helped by one tool that, unfortunately, works only on NTFS file systems (fortunately, the RAM disk itself can be in the FAT-32 system: it is faster, and reliability and fault tolerance in case of failures are not so important to us). This tool is NTFS junction points . Primo Ramdisk (VSuite Ramdisk II) has a built-in utility for creating them, but you can also use the Junction utility (and starting with Windows Vista, you can simply use the mklink command). The algorithm is very simple: copy the folder to the RAM disk (you can reassign it there as you like), for the reserve, rename the original folder to something like “original_name_bak” and create a connection point in the parent folder of the original with the original name. Everything works transparently, the program will not make any difference between the previous work with the originals and the new redirection to the copies.
Finally, it should be noted that when adding dictionaries and updating indexes, you need to synchronize the contents of the folders in the memory and backup originals on the disks - in case if the system fails, the program that manages the RAM disks will not have time to reset the contents to the images or the images will be damaged. Then you can always be restored from backup.
With full-fledged movement, the program starts and closes, synchronizes the word list with the word entry field, switches languages, searches for words and opens cards, but especially dramatically — many times — full-text search and mask search in headers accelerate.
***
Work with a RAM disk should also be adopted by those who digitize dictionaries. Sometimes you have to work with gigabyte text files. And although there are text editors for working with such huge sources, search, manual editing and autochange operations (used all the time when creating DSL language markup) occur very slowly. If you put the source into memory, and editorial work with it will be accelerated, and the compilation may also go faster.
All these skills can be applied to other programs. Since translators use the browser now as often as dictionaries, it’s worth thinking about moving browser profiles with their databases and browser caches to memory. Fortunately, there are many articles about this, even on Habré alone.
Among the electronic dictionaries, ABBYY Lingvo is distinguished by one key feature: full-text search with indexing. Something similar can be implemented with the help of indexes in Adobe Acrobat, but the convenience of the interfaces in the dictionary field is beyond comparison.
ABBYY Lingvo has long been transformed from a regular dictionary into a universal source aggregator. In addition to the creator’s titanic work, enthusiasts digitized hundreds of manuals in Lingvo format, including basic bilingual dictionaries, and multi-volume dictionaries of the Cambridge, Collins, Longman, Merriam-Webster, Oxford, and encyclopaedic dictionaries like Britannica. Local copies of network giants have been created (Wikipedia, Wiktionaries, Urban Dictionary, and so on). And in normal use, this would provide exceptional opportunities. But with full-text search, all this wealth turns into language corpuses and correspondence bases. The value of such a search when translating complex terms, stable phrases, phraseological units is difficult to overestimate.
')
With each release, ABBYY expands the allowable limits of compiled dictionaries and search indexes. Already, you can compile a dictionary of almost 2 gigabytes of source text. However, when connecting a large number of dictionaries, the index grows. Both the dictionaries on the disk and the user search index can also reach gigabyte sizes. In this case, full-text search slows down, the speed of work of the hard drives starts to influence it. The era of SSD can help in solving this problem, but so far these mechanisms are not used everywhere because of the higher price and less wear resistance. Fortunately, there is a way that wins even speed SSDs.
You can force the program to move the dictionaries and search indexes themselves from disk to memory. Rather, we ourselves will move them there, and the program will continue to work with them as if they are still on the hard drive. The old proven tool will help us - a RAM disk.
There are many programs that create the similarity of a disk in memory. Special attention should be paid to those that allow the use of memory, inaccessible to other needs.
In modern 32-bit non-server Windows, there are two main limits of memory usage: software restriction to four gigabytes , which cuts down to three memory allocation features for devices . For these reasons, no matter how much memory you put, the main Windows user will not see more than 3 gigabytes plus / minus a little more.
However, these limitations have their own workarounds: PAE and memory remapping (memory hole, PCI hole ). Therefore, if you have a non-64-bit or non-server system and at the same time three gigabytes of memory is not enough to accommodate a RAM disk with everything you need, you will have to turn on all these things and increase the memory. It is better to check the BIOS yourself: for example, in this useful article certain devices are denied support for the described workarounds, however I found the “memory hole” option in the bios of my old doomed motherboard, after switching on which everything worked as it should.
Before all the described manipulations, it is useful to archive important data and create an up-to-date image of the system for recovery if something goes wrong.
To date, there are two well-known programs that allow you to use all the invisible memory for the system: Primo Ramdisk (VSuite Ramdisk II) and SuperSpeed ​​RamDisk Plus . If anyone plans to do with three gigabytes, you can pay attention to the very fast (and according to some data - see here and here - the fastest) RAMDisk Enterprise . All these programs allow you to dump the contents of RAM disks into images on the hard drive before shutting down the computer or restarting and restoring the images to memory at the earliest stage of the system boot. Although this will slow down the download itself, the future gain is worth even a minute delay at launch.
Most motherboards for desktops and laptops of the last 6-7 years allow you to install up to 8 gigabytes of memory. Now it is a fairly inexpensive pleasure, and this size should be enough even for the richest set of dictionaries in Lingvo. However, in extreme cases, you can get by with a smaller size.
What can be moved to Lingvo on a RAM disk?
1. The simplest is dictionaries, both user and system. The translator connects user dictionaries by itself, they can be located anywhere. System dictionaries on Windows XP are located in the folder c: \ Documents and Settings \ All Users \ Application Data \ ABBYY \ Lingvo \ version number \ Dic \ System.
Both groups of dictionaries can be left in their original place for backup, and the program can point to a copy folder using dictionary configuration files. For user dictionaries, this file is% USERPROFILE% \ Local Settings \ Application Data \ ABBYY \ Lingvo \ version_name \ Dic \ dictconf.ini, for system dictionaries - c: \ Documents and Settings \ All Users \ Application Data \ ABBYY \ Lingvo \ version_DiCy \ Dic \ dictconf.ini. In both files, it is necessary to replace the old addresses with new ones with the help of autochange, before not forgetting to save backup copies of both ini-files. If this method seems inconvenient, you can resort to the following.
Naturally, all the manipulations proposed in this and the following paragraphs should be done after closing the program and unloading it from memory (Lingvo can complete work for a long time, check better for programs that display the activity of processes).
2. You can move the folder with the program and binary libraries (c: \ Program Files \ ABBYY Lingvo version_number). Perhaps this will not give an impressive increase in speed, but something can be won. Moreover, this folder also contains numerous .amd and .amm files responsible for morphology. If the program itself does not load them into memory completely, but repeatedly loads from disk if necessary, then the movement makes sense.
3. And finally - vocabulary indexes. For system dictionaries, this will be the c: \ Documents and Settings \ All Users \ Application Data \ ABBYY \ Lingvo \ version_name \ Index folder, and the user’s directory is% USERPROFILE% \ Local Settings \ Application Data \ ABBYY \ Lingvo \ version_Number \ NonAbbyyIndex (although move the entire parent folder; for system dictionaries it is better not to do this, because along with the index there are sounds in the parent folder that weigh a lot and are rarely needed).
There should be another difficulty. If moving dictionaries can be supported by editing configuration files, then moving indexes and the program itself is more intricate. But here we will be helped by one tool that, unfortunately, works only on NTFS file systems (fortunately, the RAM disk itself can be in the FAT-32 system: it is faster, and reliability and fault tolerance in case of failures are not so important to us). This tool is NTFS junction points . Primo Ramdisk (VSuite Ramdisk II) has a built-in utility for creating them, but you can also use the Junction utility (and starting with Windows Vista, you can simply use the mklink command). The algorithm is very simple: copy the folder to the RAM disk (you can reassign it there as you like), for the reserve, rename the original folder to something like “original_name_bak” and create a connection point in the parent folder of the original with the original name. Everything works transparently, the program will not make any difference between the previous work with the originals and the new redirection to the copies.
Finally, it should be noted that when adding dictionaries and updating indexes, you need to synchronize the contents of the folders in the memory and backup originals on the disks - in case if the system fails, the program that manages the RAM disks will not have time to reset the contents to the images or the images will be damaged. Then you can always be restored from backup.
With full-fledged movement, the program starts and closes, synchronizes the word list with the word entry field, switches languages, searches for words and opens cards, but especially dramatically — many times — full-text search and mask search in headers accelerate.
***
Work with a RAM disk should also be adopted by those who digitize dictionaries. Sometimes you have to work with gigabyte text files. And although there are text editors for working with such huge sources, search, manual editing and autochange operations (used all the time when creating DSL language markup) occur very slowly. If you put the source into memory, and editorial work with it will be accelerated, and the compilation may also go faster.
All these skills can be applied to other programs. Since translators use the browser now as often as dictionaries, it’s worth thinking about moving browser profiles with their databases and browser caches to memory. Fortunately, there are many articles about this, even on Habré alone.
Source: https://habr.com/ru/post/150810/
All Articles