The Splitting view allows to further split the text into snippets. This is very useful to partition large chunks of texts into smaller parts.
A very common situation is doing feedback analysis, where typically a piece of feedback contains multiple opinions, each of them potentially having different sentiments and topics. To have a granular analysis, it is important to isolate the sentiment of each opinion (positive, negative or neutral) and attach the right topics to each opinion.
The splitting view has a way to turn on/off splitting when needed, allowing you to choose how to split the text with a quick preview of the results:
You can also define filters that will run on top of the extracted snippets. This allows to further refine and remove noise from snippets and it works in a similar fashion to the cleaning view filters.
If you Save and Split, the splitting settings will be saved and all the dataset will be processed. If you go back to the dataset view, you will see that now each row corresponds to a snippet, instead of the original text. Models will be trained on top of snippets, instead of the original text.