Ivo Benke

Status: Open

Problem Description

Emotion-salient agents or affective chatbots are receiving interest these days in practice and research. Such affective chatbots rely on the domain of affective computing in order to support humans and are concerned with the design and development of computer systems that sense, interpret, adapt, and potentially respond appropriately to human emotions. One potential use-case for affective chatbots is emotional support of in virtual team communication. To stay as at least immersive as possible text-based communication over tools like Slack or Microsoft Teams, sentiment analysis und facial expression analysis over web-cam present themselves as modalities for emotion recognition. Here already tools and packages exist to track emotions on the web. Emotions recognition in conversations need to consider further aspects like utterances and specific turn. To interpret emotions right a processing and combination of the results needs to be conducted, a so-called fusion of multiple modalities.

Taken these aspects, based on recent literature in the domain of affective computing, the goal of this thesis is to develop a fusion model for the emotion recognition and processing in order to create affective chatbots. As second step this model shall be prototypically implemented into an affective chatbot and evaluated in a first pre-test.

The scope of this work can be discussed and adapted to the individual capabilities and requirements.

Goals of this Thesis

• Deriving an overview of modality collection and fusion models
• Development of fusion model for emotion recognition within textbased communication
• Prototypical implementation into affective chatbot
• Lightweight evaluation with PreTest.

Skills Required

• Knowledge in Python, Javascript and node.js or willingness to improve.
• General interest in emotion recognition, chatbots and communication.
• Reading and text analysis skills
• Very good English skills
• Very good time management and organizational skills