Memories, termbases and matches

A new project with an English-to-French Word document, auto-sectionized. Two translation memories are created and filled from TMX files, one writable and one read-only reference TM. A termbase is created and filled from a MultiTerm export, and two regex libraries for the pair are attached, one with Error Checking.

1. Create a project and choose No anonymization needed

   Create a project for the job. This document is published material, so choose No anonymization needed.

   

2. Import the Word document first

   Import the Word document first. The project takes its language pair from it, and the libraries offered later are matched to that pair.

   

3. Check the file in Project Files (Auto Sectionize split it at headings)

   The document is in Project Files, English to French. Auto Sectionize split it at its headings.

   

4. Open TMs, click New and name the TM

   Open TMs and click New. Name the translation memory.

   

5. Tick Write to this TM; set Max stored context window

   Allow Write to this TM. Max stored context window keeps the segments around each entry, so a match in the same context can score above 100%.

   

6. Tick the TM, click Import TMX and choose the file

   Tick the new TM, click Import TMX and choose the TMX file. Its entries go into the ticked TM.

   

7. Check the TM holds the entries

   The status bar confirms the import, and the TM now holds all 24.

   

8. Add a second, reference TM with Write to this TM unticked

   Add a second TM the same way: here a reference TM shared across jobs. Click New, name it, and leave Write to this TM unticked, so your confirmed segments are not saved to it.

   

9. Tick only the new TM and click Import TMX

   Untick the first TM and tick the new one: Import TMX fills the one ticked TM. Choose the reference TMX.

   

10. Leave Read on for both TMs

    Both TMs have Read on: the editor looks up both and lists the best matches first, from either TM. Confirmed segments go only to TMs with Write on, so the reference TM stays as shared. The tick boxes only pick TMs for the toolbar buttons.

    

11. Open Termbases, click New, name it and set the language pair

    Open Termbases and click New. Name the termbase and give it the project's language pair.

    

12. Tick the termbase and click Import MultiTerm XML

    Tick the termbase, click Import MultiTerm XML and choose the exported termbase. The termbases guide covers MultiTerm import and export in full.

    

13. Click the EN to FR cell

    bAIbel AV lists the languages in the file. Click the EN to FR cell to import English terms with their French equivalents.

    

14. Pick the termbase and click Import

    Pick the termbase you created and click Import.

    

15. Check the termbase holds the term pairs

    The termbase holds 26 term pairs. Read is on, so its terms are looked up in every segment.

    

16. Open Useful Regexes, click Attach library and tick the libraries

    Open Useful Regexes and click Attach library. Only libraries for the project's pair are offered. Tick both English to French libraries.

    

17. Tick Error Checking on the numbers and dates library

    Read shows each match in the editor with its French form. Tick Error Checking for the numbers and dates library: a segment then cannot be confirmed until the French follows its rules. The hints library only highlights.

    

The editor shows TM scores (including 101% context matches), termbase hits and regex matches. A 96% match is inserted and corrected, a word made bold and the segment confirmed. A link code is left out on purpose: export is refused and the log names the missing codes. F8 finds the segment, the codes are placed with Alt+1, and the export goes through.

1. Open the document and press Ctrl+Shift+E to show Termbase Hits and TM Results

   Open the document. The match column shows each segment's best TM score: segment 3 scores 101%, because its neighbour matches too. Press Ctrl+Shift+E to show the Termbase Hits and TM Results panels.

   

2. Look at segment 4: a 96% match, termbase terms and a regex hit

   Segment 4 has a 96% match: one word differs. On the left, the termbase gives moisture content and moisture meter, and the regex library flags 20 °C.

   

3. Point at a match's score to see which TM it came from

   Segment 16 has a 100% match from the reference TM. The number beside each match is drawn in its TM's colour, the swatch in the Colour column of the TMs list. Point at the number to see which TM the match came from.

   

4. Click a regex match to insert it

   In segment 11 the library finds the date, the month name and the standard number, each with the form French expects. Click a match to insert it.

   

5. Click Formatting View to see bold, italics and codes

   Click the Formatting View button. The source now shows its formatting: bold, italics, superscript, subscript, and codes for the link and the footnote.

   

6. Place Codes: choose segments and all or mandatory codes

   Place Codes has a language model put the codes into translations typed without them. Choose which segments it covers and whether it places all codes or only mandatory ones.

   

7. Note: Place Codes sends text to your AI provider

   Place Codes sends text to your AI provider. If a document holds personal or confidential data, you need a data processing agreement (DPA) with that provider, and a privacy profile is recommended. Privacy profiles have their own video.

   

8. Press Ctrl+1 to insert the 96% match and correct the word

   In segment 4, press Ctrl+1 to insert the 96% match. One word differs: select l’humidimètre and type un humidimètre. The match keeps the no-break space in 20 °C.

   

9. Select a word, press Ctrl+B, then Ctrl+Enter to confirm

   Select étalonné and press Ctrl+B, as in the source (Ctrl+I and Ctrl+U work the same way). Then press Ctrl+Enter: the regex check passes, and the segment is confirmed and written to the TM.

   

10. Leave a link out: the yellow strip lists the missing codes

    The editor moves on to segment 12, the next with codes to place. It has a bold word and a link. Type it and make chaque bold, but leave the link out. The yellow strip lists the codes still missing, and a red triangle marks the segment.

    

11. Try Export File: AV refuses; click Log to see why

    Try to export now, with Sync and Exit and Export File, and choose Leave blank for the untranslated segments. AV refuses: the export is blocked while a mandatory code is missing. Click Log: the notification log names segment 12 and the two codes it lacks, the link and the formatting inside it. AV also saves an errors report in the project folder. Nothing is exported until the codes are back.

    

12. From the top, press F8 to find the segment with missing codes

    Before you export, check the codes. From the top, press F8: it jumps to the next segment with a mandatory code missing.

    

13. Place the codes with Alt+1

    The link needs four codes. Click before page and press Alt+1 twice, then after humidité and press Alt+1 twice more. The strip empties and the triangle clears.

    

14. Click Sync and Exit, right-click the file and choose Export File

    Close the editor with Sync and Exit. Right-click the file and choose Export File.

    

15. Choose Leave blank: the export goes ahead

    The other segments have no translation yet. Leave blank exports them empty. This time no mandatory code is missing, so the export goes ahead.

    

16. Open Log to see the warnings and the file's path

    The file is exported. Its warnings are the segments left blank: the notification log lists them, with the full path of the file.