Sarba - Kompus 2017

The is a massive collection of Reddit comments designed for training and evaluating sarcasm detection systems. Release Date: April 2017.

It relies on "self-annotation," where users explicitly mark their own sarcastic comments with a /s tag, ensuring the labels are highly accurate compared to manual annotation by third parties.

You can find more detailed research on this corpus and its applications in the sarcasm identification systematic review on ResearchGate . [1704.05579] A Large Self-Annotated Corpus for Sarcasm

A sarcastic sentence might use positive words (e.g., "Great job!") to convey a negative meaning.

The "Sarba" part of your query appears to be a slight misspelling or phonetic variation of "Sarcasm" or "Sarc," the primary subject of that corpus. Overview of SARC (2017)

The way sarcasm is expressed varies across different social groups and platforms.

Identifying sarcasm is one of the hardest tasks in AI because it requires understanding: Sarcasm often depends on what was said previously.