Context:
- LM experiments (PoC: Jasmin)
- Using ChatGPT 3.5 or 4 ), make notebook available on Github
- Task 1 - use GPT as annotator0 to run 2-way annotation (annotator1 one of us) + arbitration on the 1800-2000 text we need to annotate
- Input: Useful also to validate GLs
- Give it the full guidelines as input? Maybe too long and costly
- Give definitions of labels, what we consider Israel/Palestine, other conventions, examples?
- Task2 - GPT to pre-tag texts with connotated language to help annotators be more objectives and precise
- Option1: Label whole text and could be an added column on SS
- Option2: Label specific spans/words, possibly with value like pos/neg/fact