split feature for introduction of captioning models

[x-CK-x]

captioning models beneficial for the natural language prompting, may introduce non-tag / filler words which may pose an inherent problem for the tool to label images.

split usage for captioning models

• Feature 1: Allow the tool to extrapolate valid tags & inferred tags from the captioning model while retaining their ordering from the prompt generated.
• Feature 2: Create some kind of semantic mapping from tags extracted from feature 1 to also retain the filler words et al in order to properly inject new user added tagging or natural language captioning of their own directly "in-place" to the original caption, since the mapping would allow for that and the order to be kept w.r.t. any replaced tagging. (A prompt separation strategy is needed, to isolate distinct parts of the natural language caption)
• Feature 3: Disable the Tag portion of the data curation tool as to not disrupt the functionality of the existing tagging support of the tool; while also retaining the functionality of specific features within the tool that support captions.