AquaChat¶
AquaChat is a chatbot system designed to provide adequate and context-sensitive answers to user queries. The project addresses the gap between existing AI chat assistants like Siri and ChatGPT, emphasizing that despite their capabilities, there are specific user needs and contexts where these tools are insufficient. AquaChat seeks to fill this gap by employing techniques such as keyword extraction, user knowledge modeling, user classification, and dialogue adaptation to deliver a more tailored and efficient user experience. The system also incorporates self-adaptation and sentiment analysis to refine responses dynamically based on user interaction.
It aims to address the challenges of user diversity, context-specificity, and dynamically adjusting to user feedback, ensuring a more intuitive and helpful interaction compared to traditional AI chatbots.
Key Features of AquaChat:
-
User Knowledge Modeling: AquaChat models the user's knowledge by analyzing previous interactions. This helps in adapting the conversation based on the user’s familiarity with specific topics, allowing for a more personalized experience. The system distinguishes between novice and expert users (e.g., "Who knows ARC also knows HDMI" – a way to segment users based on technical knowledge).
-
User Classification: The chatbot classifies users based on their knowledge level and preferences, allowing it to adjust its communication style and information depth. This ensures both laypeople and experts can benefit from the interaction.
-
Dialogue Adaptation: Once the user's knowledge and preferences are identified, the system adapts its dialogues accordingly. This involves simplifying responses for beginners or diving into more complex explanations for advanced users.
-
Self-Adaptive System: AquaChat employs a self-adaptive mechanism, meaning it can adjust its behavior over time based on user feedback and engagement, improving the relevance and helpfulness of responses.
You will find here more information about the AquaChat research project soon!