This page highlights recent peer‑reviewed papers, preprints, and related research artifacts produced by the Knowledge Discovery Lab in Information Retrieval, emphasizing rigorous methods and real‑world impact.
Research Produced by KDL
Cold-Start Problem in Recommender Systems
Lujie Wen, Boyang Meng, Karthik Ravi
A recommendation system suggests products, services, or content to users based on their preferences and behaviors. These systems analyze data, such as past interactions, ratings, and purchase history, to predict what a user might like or need next. The cold start problem is a common challenge in recommendation systems, machine learning models, and various other data-driven systems where a lack of initial data makes it difficult to generate accurate predictions or recommendations.
Boosting YOLO Using KAN Network – Paper
Xiaofeng Zhao
Real-time object detection technology is rapidly advancing and finding applications across various fields. YOLO (You Only Look Once), a leading framework in computer vision, continues to push the limits of performance. This project will explore the integration of the KAN Network into the YOLO framework by replacing parts of the backbone network, such as convolutional layers and activation functions. (KAN illustration by DALLE-3)
Fine Grain Sentiment Analysis – Paper
Lana Do
Fine-grained sentiment analysis goes beyond simply identifying positive or negative emotions; it captures the nuances, intensity, and specific aspects of sentiment in text. This level of detail is crucial for accurately understanding customer feedback, social media opinions, or product reviews, where mixed emotions or subtle sentiments often play a key role. By uncovering these intricacies, fine-grained analysis provides deeper insights, helping businesses and researchers make more informed decisions and create tailored responses.