With the recent introduction of so many chatbots a major trend is the feeling of disconnectivity from both the real world and the users in the system’s response. To solve this problem we want to introduce a chatbot, nicknamed Liji, with a new prompting schema, which updates its responses according to outside information. We plan to prove that by adding additional sources such as weather api and calendar information into the prompt, Liji can comprehend the user's emotional state better than other baseline chatbots, which would lead to a more specific understanding of a user and higher quality responses. Liji will give more user-friendly responses than other chatbots. For example, most people tend to feel more relaxed when they have less time than usual or the weather is nice. Do such external factors change people’s attitude toward the chatbot? And can the chatbot predict how the user feels in a certain situation? We plan to answer these questions in our research.
Blind experimental setup: To evaluate the model we plan on using a blind experimental setup with two sets of evaluation metrics, one with a in-test survey used to note the quality of the conversations with the participants, and one where participants are asked to evaluate the output of a control chat bot and Liji based on some predefined metrics (that we will define later in this proposal) and on a subject emotional connectivity metric after the each chat with the bots. We use these metrics as they are well defined in prior works and helpful to evaluate the research problem posed above. We ensure that our metrics are without bias by comparing to a control chat bot to account for any personal direction a participant would lean to.
Lorem ipsum is placeholder text commonly used in the graphic, print, and publishing industries for previewing layouts and visual mockups.
Lorem ipsum is placeholder text commonly used in the graphic, print, and publishing industries for previewing layouts and visual mockups.
- Yewon Song (Theoretical Issues, Supervisor)
- Roman Negri (Machine learning expert)
- Yuri Hwang (Writing paper)
- Kefan Xiao (Program Developer)