# Michael Beijer — full content
> Dutch patent translator and technical translator based in Hastings, UK. 30+ years specialising in patents, manuals and terminology.
---
# How to stop Trados Studio from randomly switching to overwrite mode
Published: 2026-04-09 · https://beijer.uk/blog/how_to_stop_trados_studio_from_randomly_switching_to_overwrite_mode/
> The Insert/Overwrite mode toggle has been annoying Trados Studio users for over a decade. Here's how to disable the Insert key and stop it from happening.
If you've used Trados Studio for any length of time, you've almost certainly experienced this: you're happily typing away in the editor, and suddenly your new text starts overwriting existing characters instead of being inserted. You look down at the status bar and see "OVR" staring back at you.
This is the Insert/Overwrite mode toggle – and it has been driving translators mad for well over a decade. Forum threads about it go back to Studio 2009, and it's still catching people out in 2026. The usual advice is "just press the Insert key to toggle it back", but that's cold comfort when you don't know how it got triggered in the first place, and doubly unhelpful if you're on a laptop that doesn't have a dedicated Insert key.
## Why does it happen?
The Insert key on a standard keyboard is a toggle: press it once and you switch to overwrite mode, press it again and you're back to insert mode. The problem is that the Insert key sits right next to the Delete, Home and End keys – all keys that translators use constantly when editing segments. It's very easy to clip the Insert key by accident without realising it.
On laptops, the situation is even more confusing. Many laptop keyboards combine Insert with another key (often Delete or F12) via the Fn key, and it's easy to trigger it unintentionally. On my Dell Precision 7740, for instance, Insert is hidden under Fn+F12 – not exactly intuitive.
Trados Studio itself has no setting to disable overwrite mode. The toggle is handled at the OS level, and Studio simply respects whatever mode is currently active. So the fix has to happen outside Studio.
## Immediate fix: restart Trados Studio
If you're stuck in overwrite mode right now and just want it to stop, the quickest fix is simply to restart Trados Studio. The overwrite mode state doesn't persist across restarts, so a fresh launch will put you back in insert mode. It's not elegant, but it works.
## If your Insert key doesn't work
Some translators (myself included) have previously disabled the Insert key using registry hacks or remapping tools, only to find themselves trapped in overwrite mode with no way to toggle it back. If that's you, there are two options.
### Use the on-screen keyboard
Press Win+R, type `osk` and press Enter. The Windows On-Screen Keyboard has an Insert key in the top-right area. Click it with Trados Studio focused. The on-screen keyboard generates virtual keypresses that bypass most key remaps, so it should work even if your physical Insert key is disabled.
### Send a virtual keypress via AutoHotkey
If you have [AutoHotkey](https://www.autohotkey.com/) installed, you can create a tiny script to send a virtual Insert keypress:
```ahk
SendInput "{Insert}"
ExitApp
```
Save this as a `.ahk` file and run it with Trados Studio in the foreground. This sends a software-level Insert keypress that bypasses registry-level scancode remaps.
## The permanent fix: disable the Insert key
Once you're out of overwrite mode, the real question is how to stop it from ever happening again. The answer is to disable the Insert key at the operating system level. Here are three ways to do it.
### Option 1: AutoHotkey (recommended for power users)
If you already use [AutoHotkey](https://www.autohotkey.com/), this is the easiest approach. Add one line to your existing script:
**AutoHotkey v2:**
```ahk
Insert::Return
```
**AutoHotkey v1:**
```ahk
Insert::return
```
This swallows the Insert key press so it never reaches any application. Save and reload your script, and overwrite mode will never bother you again. This is my preferred method – it's transparent, instantly reversible, and you can still send a virtual Insert keypress from a script if you ever need to (since AHK intercepts at a different level than `SendInput`).
### Option 2: Microsoft PowerToys (recommended for most users)
[PowerToys](https://learn.microsoft.com/en-us/windows/powertoys/) is a free set of utilities from Microsoft. Its Keyboard Manager module lets you remap or disable any key through a simple GUI – no scripting needed.
1. Install PowerToys from the Microsoft Store or from GitHub.
2. Open PowerToys and go to **Keyboard Manager** in the sidebar.
3. Make sure **Enable Keyboard Manager** is toggled on.
4. Click **Remap a key**.
5. Click the **+** button to add a new mapping.
6. In the **Select** column, choose **Insert**.
7. In the **To send** column, choose **Disable** (or leave it empty/undefined).
8. Click **OK** to save.
The Insert key is now dead for as long as PowerToys is running. PowerToys starts with Windows by default, so this is effectively permanent.
### Option 3: SharpKeys (registry-level remap)
[SharpKeys](https://github.com/randyrants/sharpkeys) is a free tool that writes a scancode remap directly to the Windows registry. Unlike AutoHotkey and PowerToys, it doesn't need to be running in the background – the remap persists at the driver level after a reboot.
1. Download and install SharpKeys.
2. Click **Add**.
3. In the left column ("Map this key"), select **Special: Insert**.
4. In the right column ("To this key"), select **Turn Key Off**.
5. Click **OK**, then **Write to Registry**.
6. Restart your computer.
The Insert key is now permanently disabled until you remove the mapping in SharpKeys.
**A word of caution with SharpKeys:** because it operates at the registry level, it can be hard to troubleshoot later. I once disabled my Insert key this way, forgot about it, and then couldn't figure out why Fn+F12 on my laptop did nothing when I needed to toggle overwrite mode back. If you go this route, make a note somewhere that you've done it.
## Which method should I use?
For most translators, **PowerToys** is the best balance of simplicity and flexibility. It's an official Microsoft tool, it has a friendly interface, and you can undo the change in seconds.
If you already use **AutoHotkey**, the one-liner is the obvious choice – and it's the most transparent option since you can see exactly what it does in your script.
**SharpKeys** works well as a set-and-forget solution, but bear in mind that you might forget you've done it and confuse yourself later (speaking from experience).
Whichever method you choose, you can finally stop worrying about your text being eaten by overwrite mode mid-segment. One less thing to break your flow.
---
# Trados Studio: soft returns (↵) vs paragraph marks (¶) in Excel segments – a display quirk, not a bug
Published: 2026-03-30 · https://beijer.uk/blog/trados_soft_returns_vs_paragraph_marks_excel/
> If Trados Studio shows paragraph marks (¶) instead of soft returns (↵) in the target of translated Excel segments, don't panic – the underlying data and the generated target file are correct. Here's what's actually happening.
If you translate Excel files in Trados Studio and have noticed that the **source** segment shows soft return arrows (↵) at line breaks, but the **target** shows paragraph marks (¶) for the exact same line breaks – you might reasonably assume something has gone wrong with your translation. I certainly did.
Here's what it looks like – the source on the left has ↵ symbols, the target on the right has ¶ symbols:
This happens when the target text is written programmatically – whether by a plugin, a batch task, or the Trados API – rather than being typed manually in the editor. I spent an embarrassing amount of time debugging this in [Supervertaler for Trados](https://supervertaler.com/trados.html) before realising it's not actually a problem.
## What's going on
Excel files store in-cell line breaks as literal newline characters. When Trados parses the source SDLXLIFF, it loads these newlines from the original file and renders them with the ↵ symbol (soft return). So far, so good.
When a plugin or batch task writes the translation into the target segment – using the Trados SDK's `ProcessSegmentPair` or similar API calls – the exact same newline character ends up in the target. But Trados renders it with the ¶ symbol (paragraph mark) instead.
The symbols are different. The underlying data is not.
## Proof: the SDLXLIFF and the exported file are identical
If you open the exported SDLXLIFF file in a text editor, both the source and target segments contain identical literal newlines. Here's a simplified excerpt – notice how the line breaks in the source and target are in exactly the same positions:
**Source:**
```
For all relevant sites, the risk analysis comprises:↵
- Identification of risks↵
- Probability of occurrence and detection↵
- Consequences/categorisation↵
- evaluation on the basis of set and adequate criteria↵
- Prioritisation
```
**Target:**
```
Voor alle relevante locaties omvat de risicoanalyse:↵
- Identificatie van risico's↵
- Kans op voorkomen en detectie↵
- Gevolgen/categorisering↵
- beoordeling op basis van vastgestelde en adequate criteria↵
- Prioritering
```
The `↵` markers show where the literal newline characters are – and they're in exactly the same positions in both source and target. The SDLXLIFF data is identical; only Trados's editor renders the symbols differently.
And if you generate the target Excel file (**Batch Tasks > Generate Target Translation**), the line breaks in the output are correct:
You can also verify this by selecting the source and target text in Trados (right-click > Select All, then Copy) and pasting both into a text editor – they'll be identical.
## Why does Trados render them differently?
Trados's editor distinguishes between newlines that were loaded by the file type filter (from the original document) and newlines that were written through the API. The former get the ↵ treatment, the latter get ¶. This is purely a rendering decision in the editor's display layer – it doesn't reflect any difference in the actual segment data or in what gets written to the target file.
This applies to Excel, Visio, and other file formats that store line breaks as literal text characters rather than as separate XML placeholder tags (the way Word documents use `` elements).
## TL;DR
If you see ¶ in your translated Excel targets where the source has ↵, and the translation was written by a plugin or batch task:
1. **Don't panic** – the data is correct
2. **Generate the target file** – the Excel output will have proper in-cell line breaks
3. **Verify in SDLXLIFF** if you want to be sure – the source and target will have identical newline characters
It's a cosmetic display difference in Trados Studio's editor, not a data integrity issue.
---
# Trados Studio 2024: how to fix the fuzzy match diff highlighting glitch in the Translation Results pane
Published: 2026-03-28 · https://beijer.uk/blog/trados_fuzzy_match_diff_highlighting_fix/
> If coloured diff highlighting in Trados Studio 2024's Translation Results pane has stopped working or is garbling text, here's the reliable fix – delete two settings folders and restore your user profile.
If you use Trados Studio 2024 and have noticed that the coloured diff highlighting in the Translation Results pane has stopped working for fuzzy matches – or worse, that text from the TM source and your active segment is being garbled and concatenated together (think "ThisThe decision" instead of showing the two versions cleanly) – you're not alone. This is a recurring bug that has been reported by multiple users on the RWS Community forum, and it has a reliable fix.
Here's what the Translation Results pane looks like when it's working correctly – additions are shown in green, deletions in red with strikethrough:
## What the bug looks like
When the bug is active, all of that diff highlighting disappears. The Translation Results pane either:
- Shows no coloured highlighting at all – the differences between the TM match and your active segment are simply not marked, or
- Concatenates/overlays the TM source text with your current segment text, producing garbled output where the old and new text run together (think "ThisThe decision" instead of showing the two versions cleanly)
The underlying TM data is fine – if you double-click a match to open the Edit Translation Unit dialog, the correct text is shown. It's purely a rendering glitch in the preview pane.
This has been discussed on the RWS Community forum in a thread titled ["Differences in fuzzy matches not displayed correctly in Trados Studio 2024"](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/studio/57904/differences-in-fuzzy-matches-not-displayed-correctly-in-trados-studio-2024), where multiple users reported the same problem. The fix that worked for them, and that I can now confirm also works for me, involves deleting Trados Studio's settings folders and restoring your user profile.
I also started my own thread about this (["Longstanding fuzzy match diff highlighting bug not yet fixed in Trados Studio 2026 beta?"](https://community.rws.com/product-groups/trados-portfolio/beta-groups/studio-beta/f/forum/61354/longstanding-fuzzy-match-diff-highlighting-bug-not-yet-fixed-in-trados-studio-2026-beta)), where we are trying to solve the issue in Trados Studio 2026.
## The fix
**Important:** Before you start, make sure you have a backup of your `.sdlprofile` file. If you don't know where it is, you can export it first via **File > Setup > Manage User Profiles**.
**Step 1 – Close Trados Studio completely.**
**Step 2 – Delete these two folders:**
```
C:\Users\\AppData\Roaming\Trados\Trados Studio\Studio18
C:\Users\\AppData\Roaming\Trados\Trados Studio\18
```
You can get there quickly by pressing `Win + R` and typing `%appdata%\Trados\Trados Studio\` – the two folders will be visible right there.
**Step 3 – Start Trados Studio.**
It will prompt you to select a user profile. Choose the default profile for now and let it finish loading. The diff highlighting should be working again at this point.
**Step 4 – Restore your user profile.**
Go to **File > Setup > Manage User Profiles > Change user profile**, then:
- Select either of the two profile options depending on your needs
- Click **Next**
- Browse to your backed-up `.sdlprofile` file
Your plugins and customisations will be restored.
## You may need to reinstall some plugins
Trados Studio stores plugins in three locations, and only one of them is affected by this fix:
- **`%AppData%\Trados\Trados Studio\18\Plugins\`** – this folder sits inside the `18\` directory you just deleted, so any plugins installed here **will need to be reinstalled**. (The `Unpacked\` subfolder regenerates automatically from the packages, so you only need to worry about the `.sdlplugin` files in the `Plugins\Packages\` folder.)
- **`%LocalAppData%\Trados\Trados Studio\18\Plugins\`** – not affected. Plugins installed here (including Supervertaler for Trados) will still be there.
- **`%ProgramData%\Trados\Trados Studio\18\Plugins\`** – not affected. System-wide plugins installed here will still be there.
If a plugin you rely on has disappeared after the fix, check which location it was originally installed to – if it was the Roaming folder, you'll need to reinstall it from the `.sdlplugin` file or the RWS AppStore.
## A word of warning
This fix works, but it's not permanent – at least not for me. The bug has come back on multiple occasions, which suggests it's a deeper issue with how Trados Studio manages its settings state rather than a simple one-off corruption. Each time it has reappeared, the same steps above have resolved it. I now keep my `.sdlprofile` backup in a known location and the fix takes about two minutes once you know the drill.
Hopefully RWS will address the root cause properly in a future cumulative update. In the meantime, if this post saves you the head-scratching, job done.
---
# Supervertaler for Trados – Getting Started screencast
Published: 2026-03-25 · https://beijer.uk/blog/supervertaler-for-trados-getting-started-screencast/
> A 16-minute screencast covering all the basics of Supervertaler for Trados: TermLens, prompt generation, AI translation, the Chat window, and more.
I've just published my first screencast for Supervertaler for Trados. It's a 16-minute walkthrough covering all the basics you need to get started with the plug-in.
In the video, I go through:
- The **TermLens** window and how it works
- How to **generate a translation prompt**
- How to **translate your project** with AI
- The **Chat window** and what you can do with it
- The various **Supervertaler websites** and help documentation
- How to **buy the plug-in**
If you've been curious about Supervertaler or just want to see it in action before trying it yourself, this is a good place to start.
## Links
- [Supervertaler for Trados](https://supervertaler.com/trados) – product page
- [RWS App Store](https://appstore.rws.com/plugin/432) – install directly from the App Store
- [Help documentation](https://supervertaler.gitbook.io/trados) – full user guide
---
# Using Talon Voice and Wispr Flow at the same time
Published: 2026-03-22 · https://beijer.uk/blog/using-talon-and-wispr-flow-together/
> How to use Talon voice commands alongside Wispr Flow dictation without them fighting over your microphone.
If you use [Talon Voice](https://talonvoice.com/) for hands-free computer control, you may have wondered whether you can dictate text with another tool at the same time. The answer is yes — with a small trick.
## The problem
Talon listens to everything you say. Even in command mode (where it only responds to defined commands, not free dictation), it still has the microphone open and will occasionally misinterpret your speech as a command. If you're dictating a translation with [Wispr Flow](https://www.wispr.com/) and you say something that happens to sound like a Talon command, things get messy fast.
What we need is a way to put Talon to sleep whenever Wispr Flow is active, and wake it up again when we're done dictating.
## The solution
Wispr Flow uses a push-to-talk model: you hold down a key combination to dictate, and it stops listening when you let go. In my case, that's `Ctrl+Win`. The idea is simple — while that key combo is held down, Talon sleeps. When you release it, Talon wakes back up.
Talon's `.talon` key bindings don't reliably capture modifier-only key combos, so we use a small Python script that polls the key state directly via the Windows API.
Create a file called `wispr_flow_ptt.py` in your Talon user folder (e.g. `%APPDATA%\Talon\user\`) with the following content:
```python
"""Sleep Talon while Ctrl+Win is held (Wispr Flow PTT)."""
from talon import cron, actions
import ctypes
user32 = ctypes.windll.user32
VK_LWIN = 0x5B
VK_RWIN = 0x5C
VK_CONTROL = 0x11
is_held = False
def poll():
global is_held
ctrl = user32.GetAsyncKeyState(VK_CONTROL) & 0x8000
win = (user32.GetAsyncKeyState(VK_LWIN) & 0x8000) or (user32.GetAsyncKeyState(VK_RWIN) & 0x8000)
both_down = bool(ctrl and win)
if both_down and not is_held:
is_held = True
actions.speech.disable()
elif not both_down and is_held:
is_held = False
actions.speech.enable()
cron.interval("50ms", poll)
```
That's it. No other changes needed.
## How it works
The script checks 20 times per second whether `Ctrl` and `Win` are being held down. When they are, it calls `speech.disable()` to put Talon to sleep. When you let go, it calls `speech.enable()` to wake Talon back up. The CPU overhead is negligible.
## A few notes
- **Windows only.** The script uses `ctypes.windll`, which is a Windows API. Mac and Linux users would need a different approach.
- **Adjust the key combo if needed.** If your Wispr Flow PTT is set to something other than `Ctrl+Win`, you'll need to change the [virtual key codes](https://learn.microsoft.com/en-us/windows/win32/inputdev/virtual-key-codes) in the script.
- **Talon commands still work** the moment you release the key. There's no noticeable delay — 50ms is imperceptible.
## Update: single-key push-to-talk with AutoHotkey
After using `Ctrl+Win` for a couple of hours, I noticed it was already causing RSI in my left hand — holding down two modifier keys repeatedly is not great ergonomically. I wanted to go back to using a single key: the media next key in the top right of my keyboard, which I can press comfortably with my right hand.
The problem is that Wispr Flow won't accept a single key as a push-to-talk shortcut. It insists: "Shortcut must include a modifier key or a valid mouse button."
The workaround is a tiny [AutoHotkey](https://www.autohotkey.com/) v2 script that remaps the media next key to send `Ctrl+Win`:
```ahk
#Requires AutoHotkey v2.0
*Media_Next::Send "{LWin down}{LCtrl down}"
*Media_Next up::Send "{LWin up}{LCtrl up}"
```
A bit convoluted, but now pressing (and holding) a single key with my right hand triggers Wispr Flow dictation while simultaneously putting Talon to sleep – and when I let go, Talon wakes up again, and Wispr Flow stops listening.
## Why not just use Talon's built-in dictation?
Talon does have a dictation mode, but dedicated dictation tools like Wispr Flow tend to be better at producing natural prose. Wispr Flow uses a large language model to clean up your speech in real time, which makes it particularly good for longer stretches of text. For translation work — where you're dictating in one language while your interface is in another — having a specialised dictation tool is a real advantage.
Using both tools together gives you the best of both worlds: Talon for precise, hands-free computer control, and Wispr Flow for fluid dictation.
## References
- [Talon Voice](https://talonvoice.com/) — hands-free computer control via voice commands
- [Wispr Flow](https://www.wispr.com/) — AI-powered dictation tool
---
# Trados termbase woes (finally being fixed?)
Published: 2026-03-14 · https://beijer.uk/blog/trados_termbase_woes_finally_being_fixed/
> Trados Studio's termbase system has been unreliable for years due to fragile MultiTerm interop and the old JET/Access database engine. RWS finally decoupled terminology handling in Studio 2024 SR1 – but has it actually fixed things?
Termbases not working reliably in Trados Studio is a long-standing issue. The underlying technology – MultiTerm communicating with Studio via inter-process calls, on top of the old Microsoft JET/Access database engine – has been fragile for years, and it shows up as:
- term recognition that randomly stops working mid-session
- termbases that won't accept new entries ("MultiTerm is unable to add the entry")
- general instability over long sessions, requiring termbase reorganisation or even project recreation
The [community forum history](https://community.rws.com/search?q=term%20recognition) is extensive (see below), and it's hard not to see this as a persistent "tax" on translators' time. As one long-suffering user [put it](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/studio/51825/term-recognition-does-not-work-in-certain-termbases-after-latest-update/): "This has been my experience when working with Trados for the past 10 years or so – term recognition simply stopping working, terms not being shown, completely unreliable F8 check."
### The root cause
The fundamental problem was architectural. Studio didn't handle terminology directly – it delegated everything to MultiTerm running in the background. That behind-the-scenes communication channel was brittle, and once it broke during a session, term recognition would silently fail. The underlying JET/Access database engine only made things worse, adding its own layer of instability.
### The promised fix: Studio 2024 SR1
In mid-2025, Daniel Brockmann (Principal Product Manager for Trados Studio) [acknowledged the problem publicly](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/multiterm/55704/unable-to-add-a-new-term-to-termbase-multiterm---error-message) on the RWS Community, responding directly to complaints about termbase reliability:
> "You are right that this has been a challenge for us for many years (no beating around the bush here). The potentially interesting news is that for the upcoming Studio 2024 SR1 release, we have done some very deep changes in our code to bring all of the terminology logic into the Studio code base itself (most of it was living in the MultiTerm code base, and communication behind the scenes between Studio and MultiTerm could become fragile). As one example, this has helped us address term recognition problems where it might not work reliably anymore after a certain time of working. This should hopefully be a thing of the past now."
Studio 2024 SR1 was released in the second half of 2025 and the [official release notes](https://docs.rws.com/en-US/trados-studio-2024-sr1-1187677/changes-for-trados-studio-2024-sr1-1230624) describe the change as a "MultiTerm decoupling" – Studio now functions independently from MultiTerm for terminology handling. The [SR1 blog post](https://www.trados.com/blog/whats-new-in-trados-studio-2024-sr1/) adds that terminology management has been "modernized, with all processing now happening entirely within Studio", and that "term recognition and search now pull results from all loaded termbases more reliably".
### Has it actually worked?
The architectural change is real and substantial – they genuinely ripped out the old MultiTerm interop layer and brought terminology into Studio's own codebase. On the developer side, the legacy `Sdl.MultiTerm.TMO.Interop.dll` has been [removed entirely](https://developers.rws.com/studio-api-docs/articles/hints_tips/Update_Plugins/how_to_update_plugins_to_trados_studio_2024_sr1.html) from the Studio installation folder.
But has it fixed the day-to-day experience? The jury is still out. As recently as October 2025, users who had upgraded to SR1 were [still reporting](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/studio/59250/term-recognition-still-not-working) that term recognition only worked for very short terms, with the problem persisting across different projects and file types. And the [developer forums](https://community.rws.com/developers-more/trados-portfolio/trados-studio-developers/f/sdk_qa/58913/sdl-multiterm-tmo-interop-trados-studio-2024-sr1) show that the decoupling introduced new bugs for third-party terminology providers.
Brockmann himself was honest about this, noting that they were "still stabilising this area after making these very extensive changes". So the right foundations may now be in place – but I'll believe it when I see reliable term recognition in my own daily work. The translators who depend on Trados deserve better than "hopefully a thing of the past".
This is actually one of the reasons I built [Supervertaler for Trados](https://supervertaler.com/trados/) – a Studio plugin that sidesteps MultiTerm entirely and uses its own SQLite-powered terminology system for lookups directly in the editor. It also includes an AI chat assistant and a batch translator. If you're tired of fighting with term recognition, it might be worth a look.
### References and links
- [Trados 2022 term recognition is not working with selected term bases](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/multiterm/56115/trados-2022-term-recogntion-is-not-working-with-selected-term-bases) (Philippe Galinier)
- [Term recognition problem](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/studio/53463/term-recognition-problem) (Margus Enno)
- [Term recognition problem](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/studio/41381/term-recognition-problem) (Anthony Rudd)
- [Term recognition doesn't work](https://community.rws.com/product-groups/trados-portfolio/f/licensing/54610/term-recognition-doesn-t-work/) (Therese Rose)
- [TB not showing Terms on Term Recognition](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/multiterm/55997/tb-not-showing-terms-on-term-recognition/) (Jenifer Araújo)
- [Term recognition in Studio](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/multiterm/53829/term-recognition-in-studio/) (Alan Patrick Hynds Carr)
- [Term recognition problem: only the terms recognized in the first (default) termbase are shown in full](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/multiterm/9401/term-recognition-problem-only-the-terms-recognized-in-the-first-default-termbase-are-shown-in-full/) (Yejun Huang)
- [Unable to add a new term to termbase/Multiterm – error message](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/multiterm/55704/unable-to-add-a-new-term-to-termbase-multiterm---error-message/) (Ines de Azcarate)
- [Term search not working in Multiterm term base](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/multiterm/52286/term-search-not-working-in-multiterm-term-base) (Stéphane Clément)
- [Issue with Term Recognition not working for Cloud Termbase](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/studio/53939/issue-with-term-recognition-not-working-for-cloud-termbase) (Hye Kang Shin)
- [Trados 2022 crashes when adding terms to term base](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/studio/53177/trados-2022-crashes-when-adding-terms-to-term-base) (Markus)
- [Termbase problem](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/multiterm/50155/term-base-problem) (dawoon jeong)
- [Words in term base not recognized](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/studio/49948/words-in-term-base-not-recognized/) (Jochen Schliesser)
- [Term Base Not Pulling Results Adequately](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/multiterm/47029/term-base-not-pulling-results-adequately) (Maryse van Caloen)
- [Term recognition does not work in certain termbases after latest update](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/studio/51825/term-recognition-does-not-work-in-certain-termbases-after-latest-update/) (Pavel Tsvetkov)
- [No Term Recognition for default termbase but for other termbases](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/studio/56728/no-term-recognition-for-default-termbase-but-for-other-termbases) (Christine Eulriet)
- [Term Recognition not working](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/multiterm/33443/term-recognition-not-working/) (Patrice Philippi)
- [Term recognition not working in Trados Studio 2022](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/multiterm/42898/term-recognition-not-working-in-trados-studio-2022/) (Charlotte Kauczor)
- [TRADOS 2021 "no results available" for more than half of the segments (though they have terms from my Term Base)](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/studio/39294/trados-2021-no-results-available-for-more-than-half-of-the-segments-though-they-have-terms-from-my-term-base) (Kathryn Arsenault)
- [Term adding not working from inside Trados Studio 2024 – 18.0.0.1013](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/studio/54662/term-adding-not-working-from-inside-trados-studio-2024---18-0-0-1013) (Pavel Tsvetkov)
- [TERM RECOGNITION STILL NOT WORKING](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/studio/59250/term-recognition-still-not-working) (post-SR1, October 2025)
---
# Introducing Supervertaler: My New Multicontextual AI Translation Tool
Published: 2025-10-08 · https://beijer.uk/blog/introducing_supervertaler/
> Why I built Supervertaler, what makes it different (document awareness, visual context, TM integration), and how to get started.
Supervertaler is available on GitHub: https://github.com/michaelbeijer/Supervertaler
After months of experimenting with various AI models for translation work, I finally decided to build something that addresses a major frustration I’ve been having: most AI translation tools treat each sentence in isolation, completely ignoring the broader context of the document.
That’s why I created **Supervertaler** — a Python-based translation and proofreading tool that actually understands context the way we translators do.
### What makes Supervertaler different?
The key innovation is that Supervertaler doesn’t just throw individual sentences at an AI and hope for the best. Instead, it feeds the AI multiple layers of context:
- **Full document awareness**: the AI sees the entire document when translating each sentence, not just isolated segments.
- **Visual context**: when your text references “Fig 1A”, Supervertaler can show that image to the AI (if you provide an images folder).
- **Translation memory integration**: it checks your TM for exact matches before sending anything to the AI — saving API costs and maintaining consistency.
- **Tracked changes from memoQ**: you can import previous editing decisions from bilingual DOCX files as additional context.
- **Custom instructions**: add project-specific instructions without needing to modify code.
### Two modes: Translate and Proofread
Supervertaler has two distinct modes:
- **TRANSLATE**: translates source text while leveraging your TM for exact matches.
- **PROOFREAD**: reviews and revises existing translations against the source — useful for post-editing work or quality checks.
The input/output format is simple tab-delimited text, which makes it easy to integrate with other tools in your workflow. In proofread mode, you also get comments explaining what the AI changed and why.
### Multiple AI providers supported
Rather than locking you into one AI provider, Supervertaler supports models via APIs from:
- Anthropic (Claude)
- Google (Gemini)
- OpenAI (ChatGPT)
This makes it possible to switch between models depending on your language pair, budget, or quality preferences.
### Why I built this
As translators, we know that context is everything. A term that’s perfect in one paragraph might be completely wrong three pages later. Traditional CAT tools segment our texts and while that’s great for leveraging TMs, it often means AI assistants miss crucial context.
Supervertaler bridges that gap.
The tool automatically chunks large documents into manageable pieces that fit within AI context windows — but each chunk still has awareness of the full document context.
### Get started
Supervertaler is open source and available on GitHub. You’ll need Python and API keys for whichever AI provider(s) you want to use.
If you’re tired of AI translations that miss obvious context clues, give Supervertaler a try. And feel free to reach out at michael@beijer.uk with questions or suggestions.
Note: I also published this article on LinkedIn: https://www.linkedin.com/pulse/introducing-supervertaler-my-new-multicontextual-ai-tool-beijer-i3npe
---
Source: https://beijer.uk/introducing_supervertaler
---
# What I look for in a CAT tool: my non-negotiable features
Published: 2025-10-05 · https://beijer.uk/blog/what_i_look_for_in_a_cat_tool/
> The core features I consider non‑negotiable in a CAT tool: dual source/target selection, fast termbase workflows, mature preview, robust bilingual table I/O, and flexible segmentation.
*This post is a work in progress. If you think I missed any essential features a CAT tool should have, let me know at* [michael@beijer.uk](mailto:michael@beijer.uk)*.*
I’m also trying to implement many of these features in my own context-aware, LLM-powered translation & proofreading tool: [Supervertaler](https://github.com/michaelbeijer/Supervertaler).
## 1. Ability to select corresponding pieces of source and target text in the grid
When translating long segments, I like to move down the source and target sides, selecting corresponding pieces of text to ensure I don’t miss anything. A good CAT tool should allow you to make two different selections at the same time. This is one of those features you don’t realize you need until you try it.
See also: https://www.proz.com/forum/memoq_support/372046-no_more_new_features_for_memoq_desktop-page2.html#3072457
## 2. Quick way to set certain terms in your termbase to forbidden/preferred
A termbase becomes dramatically more useful when you can quickly mark terms as *preferred* or *forbidden* during day-to-day work.
Practical expectations for me:
- One shortcut-driven workflow to mark a term pair as preferred/forbidden.
- Clear visual feedback in the UI.
- It should be easy to undo/revert the status.
## 3. A mature source/target document preview pane
Only a very few CAT tools feature a proper document previewing system. memoQ currently has the best one I have ever used.
Here’s what I am looking for in a preview system:
- The preview pane should be detachable and resizable (including on a second monitor).
- It should be possible to snap the preview pane to any part of the main UI.
- It should be possible to increase/decrease the font size, preferably via Ctrl+scroll.
- It should be possible to toggle between source and target display in the preview.
- The preview content should auto-update as you edit.
- It should be possible to scroll the preview and click somewhere in it to jump to the relevant segment in the grid.
The memoQ preview pane features all of the above, and once you have gotten used to using these features, you can’t go back.
## 4. Ability to import/export a bilingual table format into the project (with segment locking, change tracking, etc.)
A CAT tool needs a mature import/export system, allowing me to get translations into and out of the project via a robust bilingual table format (either in `.docx` or `.rtf`).
Such a system should include:
1. The option of having the import/export system ignore any locked segments in the grid.
2. Tracked Changes, so I can see exactly what got imported into the grid during an import.
## 5. All actions (such as confirming segments, adding glossary terms, etc.) must be fast and responsive
One of the main things that makes a CAT tool feel good or bad is how long things take. How quickly routine actions are completed can make the difference between a tool that feels intuitive and a tool that feels frustrating.
For example, something that really annoys me about working with memoQ is how sluggish it can feel at times. Confirming a segment can sometimes take up to five seconds. While that may not sound significant in isolation, over the course of a large project with thousands of short segments, these delays can quickly add up—potentially doubling the time required to complete a job. Not only does this affect productivity, but the constant waiting can be emotionally draining as well.
The same principle applies to adding new entries to the termbase. In some tools – like CafeTran (or memoQ) – this operation is nearly instantaneous and requires just a simple shortcut. However, this isn’t always the case across all CAT tools. Ideally, a user should be able to select the relevant terms and add them to their glossary with a single action – quickly and without interruption.
When these core operations are fast and seamless, the CAT tool feels responsive, keeping users productive and satisfied.
## 6. The tool must allow easily switching between sentence-based and paragraph-based segmentation
When working on marketing or creative content, a rigid sentence-by-sentence translation approach can seriously undermine the natural flow of the text. Unlike technical or legal documents, marketing copy relies on rhythm, nuance and cohesion across entire paragraphs.
A good CAT tool must therefore allow you to seamlessly switch between sentence-based and paragraph-based segmentation. Forcing the text into isolated segments can lead to awkward, stilted translations that miss the mark. You need the freedom to reshape, reorder and reformulate ideas across sentences – not be trapped by them.
---
Source: https://beijer.uk/what_i_look_for_in_a_cat_tool
---
# What are .mtf, .mdf, .ldb and .log files (Trados TMs/termbases), and can I delete them?
Published: 2025-08-19 · https://beijer.uk/blog/trados_mtf_and_mdf_files/
> A practical explanation of common Trados/MultiTerm sidecar files and when it’s safe to remove them.
## MTF files
- **MTF** stands for **MultiTerm Termbase File**.
- These are temporary/supporting files created when working with SDL MultiTerm.
- They are **not** the termbase itself — the real termbase is stored in the `.sdltb` file.
- MTF files are often generated during export, import, or when a termbase is opened.
## MDF files
- MDF files are supporting database files linked to MultiTerm or (in some cases) older Trados translation memories.
- They can appear alongside other database components such as IDF, LDF or NDF.
- The actual termbase remains the `.sdltb` file, while translation memories are stored in `.sdltm` files.
## LDB files
- `.ldb` stands for **Lock Database file**.
- When an Access-based database (such as `.sdltb`) is opened, Windows creates an `.ldb` (or `.laccdb`) file to manage multi-user locking.
- It records which processes have the database open.
- It should disappear automatically once all applications using the termbase are closed.
## LOG files
- `.log` files are diagnostic/activity logs created by Trados Studio or MultiTerm during operations such as TM maintenance, termbase import/export, or error handling.
- They can be useful for troubleshooting and auditing operations.
- They do **not** contain active TM or termbase data.
## Can they be deleted?
- **MTF/MDF**: often safe to delete if you retain your primary `.sdltm` (TMs) and `.sdltb` (termbases). Trados/MultiTerm can typically regenerate them when needed.
- **LDB**: only delete if you’re sure the associated termbase is closed everywhere. If it remains after a crash/unexpected shutdown, it is usually safe to remove once no process is using the database.
- **LOG**: generally safe to delete when you no longer need them for troubleshooting or record-keeping.
As a precaution, back up any active project folders before removing files — especially in shared or multi-user environments.
## See also
- The Trados Tax: Why working with Trados Studio sucks — https://beijer.uk/trados_tax
- Update Main Translation Memories vs Update Project Translation Memory (Trados Studio) — https://beijer.uk/Update_Main_Translation_Memories_vs_Update_Project_Translation_Memory_(in_Trados_Studio)
---
Source: https://beijer.uk/Trados_MTF_and_MDF_files
---
# Why do laptop manufacturers no longer sell models with actual buttons under the touchpad?
Published: 2024-08-12 · https://beijer.uk/blog/why_do_laptop_manufacturers_no_longer_sell_models_with_actual_buttons_under_the_touchpad/
> A rant from the perspective of someone doing serious work on laptops: physical touchpad buttons, proper arrow keys, and ergonomic design still matter.
These are sad times for anyone doing serious work (coding/writing/translating) on a laptop.
Why do companies keep removing features to make things sleeker and more minimalistic when it makes the computer less ergonomic? I spend an inordinate amount of time translating on my laptop, and I need:
- a touchpad with **buttons under it**
- properly sized up/down keys in a T-shaped navigation cluster (preferably separated from the rest of the keys, not squished in between them)
Why can't they keep at least one model for us professionals?
I'm currently stuck on my old-ish Dell Precision 7740 because all Dell's newer, so-called "workstation" models now have suboptimal keyboards with messed-up arrow keys and no buttons below the touchpad.
I've used Dell workstations for around 15 years, but the 7740 will be my last Dell. It looks like I'm going to have to get a ThinkPad since Lenovo is one of the few companies left that still offers semi-decent keyboards.
I literally have £3,000 to spend on a work laptop, and Dell no longer offers a single professional model.
## References
- https://techywithyou.com/laptop-with-mouse-buttons/
- https://www.shopvenom.com/blackbook-pro-17/
- https://www.metabox.com.au/store/Professional-Mobile-Workstations
- https://www.quora.com/Why-have-laptop-designers-moved-away-from-touchpad-buttons
---
Originally posted at: [https://www.linkedin.com/feed/update/urn:li:activity:7204962417225687040/](https://www.linkedin.com/feed/update/urn:li:activity:7204962417225687040/)
---
# tranzlashion.com - my spoof translation company website
Published: 2024-06-01 · https://beijer.uk/blog/tranzlashion/
> A satirical take on the cheap, questionable translation agencies that seem to pop up everywhere. Because every single language deserves to be translated immediately!
I created [tranzlashion.com](https://tranzlashion.com) as a parody of the cheap, dodgy translation agencies that seem to pop up everywhere online.
You know the type: flashy websites promising 'lightning speed' translations at rock-bottom prices, vague claims about 'quality', stock photos of people pointing at keyboards, and a suspicious overuse of the ® symbol.
The site features all the hallmarks of a questionable LSP:
- **Meaningless slogans** — 'Because every single language deserves to be translated immediately!'
- **Impossible promises** — Get a free quote in 'light speed'!
- **Vague buzzwords** — We will 'take you by the hand and guide you to grow your language needs into the future!'
- **The customer is always right** — 'We will assure you and/or your multiple clients and customers who is the boss of quality. Yes, that's right: **you are**.'
- **Proprietary magic** — 'patent pending' tools and 'psychic intuition'
## Why I made it
After 30 years in the translation industry, I've seen countless agencies come and go. Many of them share a similar aesthetic: cheap templates, stock imagery, buzzword-heavy copy, and promises that would make any professional translator wince.
I thought it would be fun to create an over-the-top parody that captures the essence of these sites. It's meant as gentle industry humour — a knowing wink to fellow translators who've received 'exciting opportunities' from such companies offering rates that wouldn't cover a cup of coffee.
Check it out at [tranzlashion.com](https://tranzlashion.com) and see if you can spot all the red flags!
*Disclaimer: Tranzlashion® is not a real company. Please do not actually request translations. If you need quality Dutch↔English translation, you know where to find me.*
---
# ProZ.com virtual conference for International Translation Day, session: 'Machine translation panel from a freelancer perspective'.
Published: 2014-09-30 · https://beijer.uk/blog/prozcom_virtual_conference_for_international_translation_day_30_sept_2014_session_machine_translation_panel_from_a_freelancer_perspective/
> I was a panellist in the ProZ.com virtual conference for International Translation Day 2014, in the session 'Machine translation panel from a freelancer perspective'.
I was a panellist in the ProZ.com virtual conference for International Translation Day (30 Sept 2014), in the session 'Machine translation panel from a freelancer perspective'. The session was moderated by Jeff Allen and included Dimitra Anastasiou, Patrick Porter, and myself as panellists, with around 105 people watching live.
The discussion was structured as a roundtable covering four main topics: each panellist's MT workflow, how to handle confidentiality when using online MT engines, the impact of post-editing, and which language pairs and directions lend themselves best to MT.
A recording is available on the [ProZ.com conference page](http://www.proz.com/virtual-conferences/632/program/9778). (Please excuse the audio quality at the start — it worked fine during the dress rehearsal!)