From a Word-Level Dictionary to Sentence-Level Semantics: Multilingual Grievance Labelling with Contextual Models

Researchers find that multilingual grievance detection is more reliable with context-aware models than with word-matching dictionaries.

A lexicon-based Grievance Dictionary can look strong on a biased test set, but its performance drops on a non-circular benchmark.

Using the full post and sentence context helps detect harder cases, including quoted, implicit, and cross-sentence grievances.

The study introduces a new benchmark and code for evaluating grievance labeling across five languages.