School of Engineering Department of Computer Science and Engineering 123 Knowledge Discovery over Database Supervisor: WONG, Raymond Chi Wing / CSE Student: REN, Xiyu / MATH-GM Course: UROP2100, Summer With advancements in NLP and ASR, voice-based interfaces have become crucial for various applications such as chatbots, search engines, and databases. The VoiceQuerySystem is a new voice-based database querying system showcased in this demonstration, which allows users to perform operations on the databases by direct natural language questions (NLQs). Unlike existing interfaces that are based on voice instructions such as SpeakQL and EchoQuery, where the vocal inputs must be exactly in SQL languages or other predefined instructions, VoiceQuerySystem’s aim is that it enables the manipulation of data through natural languages. This feature eliminates the need for users to have technical knowledge of SQL language. Knowledge Discovery over Database Supervisor: WONG, Raymond Chi Wing / CSE Student: SU, Hong / DSCT Course: UROP4100, Fall UROP4100, Spring Recommender systems are an increasingly popular tool for improving user satisfaction and content provider revenue. However, many current recommendation models lack interpretability, making it difficult to understand why recommendations are made. Explainable recommendation models aim to address this issue by providing personalized recommendations along with explanations for why the recommendations were made. These models can be either model-intrinsic or model-agnostic and use various information sources, such as relevant user/item explanations, textual explanations, and visual explanations. While session-based recommendation is a promising research field, it is still a new problem to study explainability in sessionbased recommendations. This report presents a literature survey on related fields in explainable sessionbased recommendations and provides datasets with high potential research value in explainable sessionbased recommendations. Knowledge Discovery over Database Supervisor: WONG, Raymond Chi Wing / CSE Student: TAN, Weile / DSCT WANG, Ruida / DSCT Course: UROP1100, Fall UROP2100, Fall The topic of this project is to research a better way of decoding information in the field of Session-based Recommendation Systems. The Session-based Recommendation is a case of next-item recommendation, which aims to predict the user’s next possible click given only what the user has clicked in the same session. Unlike most of the research in the field, which mostly focuses on how to encode the information in the session, our research mainly focuses on formulating a better decoder for the information, which concerns information that previous decoders may ignore. Our alternative goal for the model is to develop an off-shelf booster that may be applied to any encoder-decoder-based model and improve their performance. The current version of our booster takes advantage of the Neural Random-Forest algorithm to enhance the users’-behavior information loss in predicting.
RkJQdWJsaXNoZXIy NDk5Njg=