I am motivated by the challenge of building systems that can extract, understand, and reason
with information in natural language texts.
Here is a listing of active research projects:
- Explainable Natural Language Inference -- This NSF supported project, under award numbers 1815358 (Stony Brook University) and 1815948 (University of Arizona), focuses on developing explainable inference algorithms. Applications include question answering, and relation extraction. This is a collaborative research effort with co-investigators Prof. Peter Jansen and Prof. Mihai Surdeanu from University of Arizona.
Our work also looks into developing datasets that explicitly discourage models from employing shortcuts and other artifact based reasoning.
- Understanding Common-sense Knowledge About Events
- Modeling conditional knowledge about events -- Events typically alter the state of the world. Reasoning about events in the real world requires understanding what enables them i.e., preconditions, and how they affect the world state i.e., postconditinos. This NSF supported project focuses on developing techniques for acquiring this kind of knowledge from news texts, and simple narratives and using them to enable applications such as story generation, planning, and answering why questions. The work will contribute large scale datasets, new structured generative models, and applications related to the technical goals.
- Modeling Schematic Event Knowledge -- This DARPA funded project focuses on developing event language models that can reason about events. In particular we focus on augmenting the event models with preconditions and postcodition knowledge, combining multimodal information from videos, and hierarchical representations of event sequences.
The work also focuses on developing generative langauge models with a particular emphasis on addressing the well-known issues of redundancy, lack of coherence and consistency.
- Efficiency and Energy Consumption in NLP -- Models are getting larger, consuming more compute, memory, and energy. Our work looks into ways for making NLP models faster, smaller, and more energy efficient where possible.
- Verifying Complex Software with NLP -- Complext software systems are hard to verify mainly because translating natural language specificaitons into formal verifiable statements is difficult. This NSF funded project is looking at developing semantic parsing systems that convert Netowrk File System specification texts into formal statements.
- Private Intelligence Assistance -- This project focuses on delivering privacy focused NLP applications that can run on user devices.