I think this is a raw (unprocessed) transcript.
Normally, the LingQ importer pre-processes transcript before importing to bond broken words (because the source is transcript-based on voice), but sometimes it doesn’t work, since this is an internal work, there’s nothing can be done about the issue, the only way is manually tidy transcript with tools.
This topic also says about this issue.
Dec/24 LingQ Changelog: AI lesson pre-processing - #15 by vet8t6z79pc4