Automatic Expansion of Domain-Specific Affective Models for Web Intelligence Applications.

Area of Science:

• Computational Linguistics
• Affective Computing
• Artificial Intelligence

Background:

• Sentic computing uses affective models for sentiment analysis and emotion recognition.
• Existing models may not capture complex, domain-specific communication goals.
• Standard affective dimensions (e.g., Joy, Trust) may not align with strategic marketing objectives.

Purpose of the Study:

• To introduce and evaluate techniques for expanding affective models.
• To improve the coverage, consistency, and domain-specific interpretation of emotions.
• To address limitations of standardized affective models in measuring communication success.

Main Methods:

• Combining knowledge graphs (common and commonsense knowledge) with language models and affective reasoning.
• Developing expansion techniques for affective models.
• Quantitative evaluation using the Hourglass of Emotions model and gold standard data.
• Qualitative evaluation of a domain-specific affective model for television brands.

Main Results:

• The proposed expansion techniques enhance affective model performance.
• The methods support various embeddings and pre-trained language models.
• Demonstrated improved coverage and consistency in affective modeling.
• Successfully adapted models for domain-specific interpretations, like in television branding.

Conclusions:

• Expansion techniques offer a robust way to adapt affective models for diverse applications.
• The approach is valuable in scenarios with limited affective model resources.
• Enables more nuanced and strategically aligned emotion analysis in communication.