Four common assumptions about Arabic dialects used in NLP are shown to oversimplify reality: dialects overlap heavily, length is a weak predictor of ambiguity, lexical cues are not distinctive, and dialectness ratings vary by annotator region.
InProceedings of the Fourth Arabic Natural Language Processing Workshop, pages 285–289, Florence, Italy
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Revisiting Common Assumptions about Arabic Dialects in NLP
Four common assumptions about Arabic dialects used in NLP are shown to oversimplify reality: dialects overlap heavily, length is a weak predictor of ambiguity, lexical cues are not distinctive, and dialectness ratings vary by annotator region.