Why does Glyssen say a passage is not aligned to the reference text, when it appears to align perfectly?

When aligning the vernacular (target language) blocks to the reference text (using “rainbow mode”) in the Identify Speaking Parts dialog box, sometimes the Verses not aligned with reference text filter will take you to a passage that appears to align perfectly. All the character cells (and Delivery cells, if those are showing) are filled in, and there is appropriate text in each reference text cell. Still, you have to click Apply in order for Glyssen to accept the alignment and consider it done. Why?

When this occurs, if you notice, in the left-hand pane, there will be one or more of the Character cells that is blank. This happens because when parsing the quotes, Glyssen encounters a verse that has two or more expected speakers and therefore marks the direct speech in that verse as “Ambiguous”. Often when this happens, the number and order of narrator and speaking blocks in the vernacular will exactly match the reference text, so if nothing else is out of alignment, Glyssen will be able to match everything up perfectly. However, there is always the slight possibility that the match is coincidental. Sometimes the vernacular may break up or re-order the speech in a way that happens to appear to match the reference text but doesn’t really. So Glyssen will take the user to these places for confirmation. If the user knows both the reference text language and the vernacular (which is the recommended practice), then it will be easy to see that they are in alignment. Once the user has made sure that the text matches properly and the character and delivery information is correct, clicking the Apply button will confirm that Glyssen’s guess was correct.

Another type of misalignment can occur when verses in the vernacular (i.e., target) language are divided up into multiple paragraphs, where the reference text has the text as a single block. This is common in poetry, genealogies, and in quoted material. In some cases, if the paragraphs are “poetry” line divisions and the lines do not end with sentence-ending punctuation, Glyssen will automatically join them into a single block to avoid this kind of misalignment. (While division of poetic material into lines and stanzas looks great on the printed page, these divisions are usually meaningless in an audio recording (unless the translator has managed to get them to correspond to rhyme or rhythm patterns).
There are two possible approaches for dealing with this kind of misalignment:

  1. Right click the reference text cell with the text and select Split Text* on the context menu. This can only be done when that cell is not in edit mode. If it is in edit mode, click away to a different cell, and then right-click the cell containing the reference text to be split up and distributed across multiple target blocks. Although there is no visual indication, you are now in “Split Text” mode. Click in the reference text where you want to split the text, and everything after that location will be moved down to the following cell. As long as there are additional empty cells below that, you will remain in Split Text mode, so you can continue to click to break up the reference text and distribute it piece by piece so that it correspond to each block in the vernacular.

  2. Glyssen will allow you to click Apply without actually aligning the vernacular blocks to the reference text. If you do not need to have the text aligned exactly, just click Apply.

With either of the above approaches, if you wish to force Glyssen to join two or more vernacular blocks when exporting the script, type at the start of the primary reference text. Any block whose primary reference text starts with will be joined with the preceding block. (If there reference text consists solely of , as would be the case if you followed the #2 approach above. that counts as “starting with” )

  • Note: Do not confuse the Split Text command on the context menu with the Split command on the main menu. The Split command is used for splitting target language blocks.