I converted the Tanaka Corpus of aligned Japanese and English sentences into Felix translation memory (TM) and TMX formats.
The Tanaka Corpus is a collection of around 150,000 Japanese-English sentence translation pairs, compiled over several years by university students, with later cleanup and correction by Jim Breen and his colleagues.
There was an interesting question on the Japanese-language Felix forum about displaying the glossary/memory name in the Felix search results (from the new search window).
Right now, you can view the memory/glossary name if you click on the “Details” link, but it’s not visible by default. I plan to make it visible in a near-future update, but in the meantime, you can modify the search-matches template to show this information in the current Felix version.
I’ve created modified versions of the English and Japanese-language versions of the templates, and made them available for download.
To install the template file, do the following.
The next time you do a search, the file name will appear in the results.
For more details about what you can customize in the template files, see Customizing Felix Templates in the manual.
I sometimes get requests about when and if I’m going to support Linux (or the Mac). In fact, several Felix users are almost completely on Linux, and only go back to Windows for applications like Felix.
I do have some non-concrete plans to support Linux (and by extension the Mac) in the future.
Firstly, I plan to offer Felix as an online service. This will essentially be Memory Serves, behind a secure server. That will make Felix itself available from any platform that supports a Web browser.
Secondly, I have a very rough OpenOffice Writer extension for Felix, with plans to improve it and add extensions for Impress and Calc. These can then be updated to work with the online service.
Thirdly, I’m currently working on an XLIFF-based translation editor, the first beta of which is scheduled for December. This is being written in Python, and it should be possible to make it run on Linux without too much extra work.
I know that many CAT tool users won’t use online services, because they’ve told me so. But some users do want a hosted service, and at any rate, as a solo developer this is the only realistic way I have of supporting platforms other than Windows.
I’ve just released version 1.3 of Memory Serves.
Download the latest version here.
This new version has greatly improved performance, as well as two new features: search and replace, and an admin interface
Memory Serves is now much faster and more responsive. In the previous versions, performance would drop markedly with very large translation memories, but version 1.3 is up to 10 times faster.
Memory Serves has a powerful new search and replace feature, making it easy to explore and maintain your translation memories and glossaries. See the section on Search and Replace in the manual for details.
A new admin interface allows you to create and manage user accounts, view logs, and configure Memory Serves preferences. See the section on the Admin Interface for details.
Memory Serves is a free program for sharing Felix translation memories and glossaries over a LAN or VPN. Using Memory Serves, a group of translators can share their TMs and glossaries in real time, ensuring consistency and quality of translations.
I’m working on the next version of Memory Serves now, so I’m doing a lot of dogfooding with it. I’ve thus been using Memory Serves pretty much exclusively over the past month for my own translation work.
Over the course of using Memory Serves intensively, I’ve uncovered quite a few areas needing improvement; which is good, because knowing about the problems makes it possible to fix them. 🙂
One feature that really saved my bacon, however, was the fact that Memory Serves keeps the database up-to-date at all times. I had been working on a fairly large translation, and went out with my family for dinner. Okinawa was experiencing some intense electrical storms, and when we got back, I found that my neighborhood had had a blackout, and my computers had all shut down.
Since Memory Serves uses the SQLite database to store the translation memories, all changes to the TMs are saved to disk immediately. So none of my work was lost, and I was able to carry on translating.
With Felix, your TMs aren’t saved automatically; you have to save them much as you would a Word document. Although it will prompt you to save if you exit the program with unsaved changes, if your computer (or Felix) crashed, then you’d lose all the translation entries you’d made since your last save.
This happened to a Felix user a few months back: she had been working on a translation for about six hours when her computer crashed, and she hadn’t saved her TM even once. She asked me if there was some way to recover her translations, but the only way was to use Align Assist to recreate her translation memory — the original TM was lost.
I added a ticket to my Felix issue tracker to add automatic background saving of TMs, but until now I’ve given higher priority to other development. Seeing first hand how this feature saved my own bacon with Memory Serves, however, I’ve decided to give it higher priority for Felix as well. I hope to have it included in Felix by the next release (version 1.5), or at the latest by the version after that (1.5.1).
The next version of Memory Serves will be released over the next few days, and it’ll have a lot of improvements as well. In particular, it’s much faster, fixes some issues with correcting/editing translations, and will have a new search and replace feature. The new search and replace will serve as a prototype of the improved search and replace I’m adding to Felix.
Built-in glossary searching is one of the key features of Felix. In this post, I want to describe how the glossary searching algorithm works, and how results are displayed.
Felix has two choices for glossary searches. If you choose a minimum score of below 100% (Tools >> Preferences >> Glossary >> “Minimum fuzzy score”), then Felix will do fuzzy matching based on the Levenshtein (edit) distance.
If you select a score of 100%, it will only count perfect matches.
To give an idea of what this means, consider this glossary entry:
aaBaa
Now, say you’re translating this sentence:
Put the aaCaa in the box.
If you’re not using fuzzy matching, then no match will be found for aaCaa. If you set the fuzzy threshold to around 80%, then this will be retrieved as a candidate.
You can also set whether to ignore case, wide/narrow character distinctions, and distinctions between Hiragana and Katakana.
Ignore…
Case: “aaa” is the same as “AAA”
Wide/narrow: “１２３” is the same as “123”
Hiragana/Katakana: “いろは” is the same as “イロハ”
All the glossary matches for the current sentence are displayed in the glossary window. The matches are displayed by reference count, string length, and score. That is, the match with the highest reference count is shown first in the list of matches; if two matches have the same reference count, then the longer match goes first; and so on.
Reference count: The number of times the translation has been retrieved by the user
String length: How long the source word/phrase is
Score: If you use fuzzy glossary matching, how close the match is.
There are several ways in which the glossary matching algorithm could be improved. Felix user Steven Venti proposed a search algorithm that I would characterize as based on “closeness” or “stickiness,” and gave the program Jamming (Japanese) as an example of a program that does dictionary searches very well.
Another feature I’ve been thinking about for a while is the ability to create rule-based glossary entries, using wildcards or regular expressions. For example, you could do this to create translations for dates, or product names consisting of set patterns.
The way that matches are displayed can also be improved. I could make it possible for users to determine the sort criteria (what order matches are displayed in), both through preferences and dynamically. I’m also planning to make it possible to easily show and hide details about glossary matches — for example, click “details” to show all the information about the match, such as creator and date created, and “minimal” to show just the source and translation (thus allowing more matches to be shown at once).
In a way, being able to specify the order in which matches are displayed could make up for the “feast or famine” problem that Steven mentions: getting either too few or too many matches. If you set the match score low enough that you get lots of matches, but could arrange so that the matches you want are shown first, I think that would go a long way toward improving usability.
I’ve just released version 1.4.7 of Felix.
Download the latest version of Felix here
Below are the main changes and improvements in this version.
Details about each improvement/bug fix follow.
In PowerPoint, the setting to skip segments if they did/didn’t contain double-byte characters was ignored.
Now if you select “Load previous memories/glossaries on startup” in the preferences, remote memories/glossaries will also be loaded. Felix will remember a total of 15 each of remote and local memories and glossaries (60 items total).
When you save your preferences, the preferences file (.fprefs) now includes information about remote memories and glossaries that were loaded. The next time you load that preferences file, Felix will connect to the same remote memories and glossaries. See the manual for details about connecting to remote memories.
The latest Trados translaton-memory text format is supported. Note, however, that text formatting information (such as bold and italic) is currently ignored in the new format. Text formatting is still supported in the old Trados text format.
It wasn’t possible to change the “locked” property of memories or glossaries if the current user’s account name didn’t match the creator value of the TM. This caused problems in some situations, like when the creator was “Align Assist.”
Felix preference files (.fprefs) are now associated with Felix. This means that if you double click on a preferences file, it will launch Felix, and Felix will be configured with all the settings and memories/glossaries in that file. You can also drag and drop an .fprefs file onto the Felix memory window, and Felix will load those preferences.
There was a bug whereby if you already had Felix installed, and then installed an update from a different account, some templates weren’t copied over to the user’s local folder.
When you import a MultiTerm glossary, the source and translation languages are set as the source and translation languages of the Felix glossary.
Before, when you launched Felix for the first time after installing it, Felix would start with the Japanese UI if you were running on a Japanese OS, and the English UI otherwise. Now, if will start in whatever language you chose for the installer (currently Japanese and English only).
I’ve just released version 1.4.6 of Felix.
Download the latest version here.
This release improves the stability of Felix in MS Word, and adds a number of new features and improvements. The main changes in Felix version 1.4.6 are as follows:
Details about each change are shown below.
Felix now includes a feature to check for updates online. Felix is updated frequently, and this feature will help you make sure you have the latest version.
The update check works in two ways. First, you can check for updates manually by selecting Help >> Check Updates… from the Felix menu.
Second, Felix will check for updates automatically every two weeks by default. It will ask you before going online to check; if you don’t want Felix to check online, click No. Select the “Don’t ask me again” checkbox if you don’t want to see this dialog again. (If you select “Don’t ask me again” and click Yes, then Felix will automatically check online for updates every two weeks, without asking first.)
If there isn’t a newer version available, Felix will let you know:
If there’s a newer version available, Felix will let you know, and ask if you want to go to the download page.
Correcting translations using the translation history feature now works with Memory Serves. You will need version 1.2.1 of Memory Serves or higher to use this feature; download the latest version here.
Each time you do a memory or glossary search, the window will scroll back to the top. This is useful if you’ve scrolled down a list of matches, and then perform a new search.
When you edit translation matches in the Search or Concordance view, and then navigate back to the translation match view (e.g. by pressing F6), the match is re-calculated to reflect any edits you’ve made.
You can press CTRL + ALT + F9 to toggle the Felix keyboard shortcuts on and off from Microsoft® Word, PowerPoint, and Excel. Starting with version 1.4.6, this setting will be saved the next time you start each program. Additionally, a minus sign in square brackets (“[-]”) appears to the right of the Felix menu name when shortcuts are disabled, as a visual cue.
In earlier versions of Felix, some users experienced instability with the Felix add-in for MS Word. Starting with version 1.4.6, the Word add-in degrades gracefully in the face of errors, instead of crashing.
When searching and replacing in Edit mode, the search dialog has the option to search/replace in source fields, translation fields, context fields, or all fields.
In previous versions of Felix, while this selection worked for the replace function, the search function always searched in all fields. This bug is fixed in version 1.4.6.
A template is now used for the glossary match view. With this, all information views in the Felix memory and glossary windows can now be customized. See the template section of the manual for details.
Entries with <, &, and > characters caused errors when getting translations as plain text. Now they’re handled properly.
In previous versions of Felix, every time you retrieved a translation in Word using the “Set and Next” command, Felix would add that translation back into the memory. This was causing problems for people who wanted to use multiple translation memories, and keep their entries separate.
Now, the Word add-in only adds a translation back into Felix if the match is not 100% (i.e. you did “Set and Next” when the match was not perfect). It adds it back in this case, because Felix assumes you’re saying that the retrieved translation is correct for this source segment.
When you import a memory in TMX format, Felix will now pick up the source and target languages, and set them as properties in your Felix memory.
Felix now uses memory more efficiently, by only loading the logging components when they’re being used. This can save up to 60 MB of RAM usage.
In previous versions, pressing Alt + C in Felix (the Concordance shortcut) with no text selected caused an error. Now, it simply displays a message that there is no text selected.
If one of your MS Office programs (Word, Excel, or PowerPoint) crashes, the Felix menu and toolbar might not appear the next time you start that program. This could be because Office has disabled the Felix add-in.
To enable the add-in, do the following.
If Felix is in the list of disabled add-ins, then set the “Manage” list at the bottom of the window to “COM Add-ins”, and click Go. Select the “Felix Addin” checkbox, and click OK.
If Felix isn’t on the list of disabled items, set the “Manage” list at the bottom to “Disabled Items”, and click Go. Enable “Felix Addin”, and click OK.
If the problem persists, please consider sending me your Felix log files. To do this:
I’ve just released version 0.9 of Align Assist. The two main features of this new release are a “Check for Updates” feature, and a link to the online manual from the application.
Download the latest version here.
From the Help menu, select Check for Updates… If the latest version of Align Assist is newer than your version, it will prompt you to download the latest version.
Align Assist will also check online for updates automatically once a week, unless you tell it not to (it will ask before going online).
From the Help menu, select Online Help. The online manual opens in your Web browser.
Align Assist is a free application that aligns legacy source and translation files, in order to make Felix translation memories (TMs).