An effective emotion tendency perception model in empathic dialogue.

Area of Science:

• Artificial Intelligence
• Natural Language Processing
• Computational Linguistics

Background:

• Open-domain dialogue systems rely on accurate emotion detection for effectiveness.
• Previous models often relied on identifying emotional words, leading to bias and imprecise quantification.
• A need exists for models that capture the emotional tendency of all words within a dialogue context.

Purpose of the Study:

• To propose a novel emotion tendency perception model for dialogue systems.
• To accurately quantify the emotional tendencies of all words, not just explicit emotional terms.
• To enhance dialogue systems by integrating sentiment and semantic understanding.

Main Methods:

• Developed an emotion encoder to precisely quantify emotional tendencies of individual words.
• Implemented a shared fusion decoder to integrate emotional and semantic information.
• Conducted extensive evaluations using the Empathetic Dialogue dataset.

Main Results:

• The proposed model demonstrated significant efficacy in emotion perception within dialogue systems.
• Experimental results showed superior performance compared to existing state-of-the-art approaches.
• The model successfully reduced bias associated with previous emotion detection methods.

Conclusions:

• The emotion tendency perception model offers a more accurate and nuanced approach to emotion detection in dialogue systems.
• This advancement can lead to more empathetic and effective human-computer interactions.
• The model's ability to fuse sentiment and semantic capabilities provides a distinct advantage in the field.