User Context

User context refers to the incorporation of historical behavioral, linguistic, and social information about authors to improve NLP tasks. On social media, where text is sparse and informal, user-level priors (e.g., sentiment tendency, sarcasm frequency, political leaning) can disambiguate ambiguous utterances.

User context is particularly valuable for tasks sensitive to author intent or style: - Sarcasm detection: Users who frequently post sarcastically are easier to identify; conversely, a generally literal user's occasional sarcasm is harder to detect - Sentiment analysis: Author sentiment priors help distinguish implicit or conflicted sentiment - Stance detection: Author political alignment provides disambiguation for ambiguous claims

Key papers

• Perceived and Intended Sarcasm Detection with Graph Attention Networks — user2vec embeddings from historical tweets improve sarcasm detection via graph attention
• The Role of User Profiles for Fake News Detection — user profiles and social relationships for fake news detection

Related topics

• Personalized NLP (broader: adapting models to individual users)
• Author Profiling (inferring demographic or stylistic properties from user data)
• Social Media Analysis (leveraging platform-specific author behavior)
• Sarcasm Detection (heavily dependent on author history and social relationships)