UCSB Natural Language Processing (NLP) Group to Collaborate with Google on Language and Vision Research
By Sandra Dieron, PR Assistant, Computer Science
The UCSB Natural Language Processing (NLP) group, directed by Computer Science Professors William Wang, are the new recipients of a Google Gift award to be used towards “Vision-Language Navigation with Self-Supervised Learning.”
Existing research studies on vision and language grounding for navigation focus on improving model-free deep reinforcement learning (DRL) models. However, model-free DRL models do not consider the dynamics in the real-world environments, and they often fail to generalize to new scenes. The sample complexity is also not ideal.
In this project, building on UCSB NLP Group's existing studies (Wang et al., 2019; Wang et al., 2018; Xiong et al., 2018) led by Ph.D. students Xin Wang and Wenhan Xiong, the NLP group will take a radical approach to bridge the gap between synthetic studies and real-world practices. The group will work with Google to improve the sample complexity of the reinforcement learning model, and further bridge the gap between model-free and model-based reinforcement learning to solve a real-world vision-language navigation task.
The UC Santa Barbara Natural Language Processing Group studies the theoretical foundation and practical algorithms for language technologies. The group tackles challenging learning and reasoning problems under uncertainty and pursues answers via studies of machine learning, deep learning, and interdisciplinary data science.
Broadly, the UCSB NLP Group are interested in designing scalable inference and learning algorithms to analyze massive datasets with complex structures. In particular, the lab concentrates in the areas of information extraction, computational social science, knowledge graph, learning to reason, dialogue systems, language & vision, summarization, statistical relational learning, reinforcement learning, structure learning, and deep learning.
Congratulations to the UCSB NLP Group on receiving a Google Gift! We are excited to see what work will be produced through this gift.