Lexicon of Polarity Shifting Directions for 2,521 Words

A list of 2,521 polarity shifters labeled for their shifting direction, derived from the lexicon of Schulder, Wiegand and Ruppenhofer (JNLE 2021). It contains 863 shifters affecting only positive polar expressions, 288 affecting only negative ones, and 1,370 that can shift in both directions. The dataset was created by Marc Schulder at Universität Hamburg for a 2020 LREC conference publication.

Use Cases

• Training classifiers to identify polarity shifters based on part-of-speech and lemma features.
• Enhancing sentiment analysis systems by incorporating rules for context-dependent polarity reversal.
• Linguistic research on the distribution and behavior of polarity-shifting words across different parts of speech.
• Validating automated polarity shifter labeling methods against a manually annotated gold standard subset.

Strengths

• Provides explicit counts for three distinct shifting direction categories: 863 AFFECTS_POSITIVE, 288 AFFECTS_NEGATIVE, and 1,370 AFFECTS_BOTH.
• Includes a source label distinguishing between a manually verified gold standard and automatically labeled entries.
• Covers a specific set of 2,521 shifters from a cited academic lexicon, providing a clear provenance.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Last update date is unknown; freshness unverified.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Lexicon from Schulder, Wiegand and Ruppenhofer (JNLE 2021), extended through supervised classification.

Collection Method

Likely contains manual annotation for a gold standard subset and automated labeling for the remainder.