Pretrained transformer-based Language Models (LMs) are well-known for their ability to achieve significant improvement on NLP tasks, but their black-box nature, which leads to a lack of interpretability, has been a major concern. My dissertation focuses on developing intrinsically interpretable models when using LMs as encoders while maintaining their superior performance via prototypical networks. I initiated my research by investigating enhancements in performance for interpretable models of sarcasm detection. My proposed approach focuses on capturing sentiment incongruity to enhance accuracy while offering instance-based explanations for the classification decisions. Later, we develop a novel white-box multi-head graph attention-based prototypical framework designed to explain the decisions of text classification models without sacrificing the accuracy of the original black-box LMs. In addition, I am working on extending the attention-based prototypical framework with contrastive learning to redesign an interpretable graph neural network for document classification, aiming to enhance both the interpretability and performance of the model in document classification.

Pretrained transformer-based Language Models (LMs) are well-known for their ability to achieve significant improvement on text classification tasks with their powerful word embeddings, but their black-box nature, which leads to a lack of interpretability, has been a major concern. In this work, we introduce GAProtoNet, a novel white-box Multi-head Graph Attention-based Prototypical Network designed to explain the decisions of text classification models built with LM encoders. In our approach, the input vector and prototypes are regarded as nodes within a graph, and we utilize multi-head graph attention to selectively construct edges between the input node and prototype nodes to learn an interpretable prototypical representation. During inference, the model makes decisions based on a linear combination of activated prototypes weighted by the attention score assigned for each prototype, allowing its choices to be transparently explained by the attention weights and the prototypes. Experiments on multiple public datasets show our approach achieves superior results without sacrificing the accuracy of the original black-box LMs. We also compare with four alternative prototypical network variations and our approach achieves the best accuracy and F1 among all. Our case study and visualization of prototype clusters also demonstrate the efficiency in explaining the decisions of black-box models built with LMs.

Point-of-Care Ultrasound (POCUS) refers to clinician-performed and interpreted ultrasonography at the patient's bedside. Interpreting these images requires a high level of expertise, which may not be available during emergencies. In this paper, we support POCUS by developing classifiers that can aid medical professionals by diagnosing whether or not a patient has pneumothorax. We decomposed the task into multiple steps, using YOLOv4 to extract relevant regions of the video and a 3D sparse coding model to represent video features. Given the difficulty in acquiring positive training videos, we trained a small-data classifier with a maximum of 15 positive and 32 negative examples. To counteract this limitation, we leveraged subject matter expert (SME) knowledge to limit the hypothesis space, thus reducing the cost of data collection. We present results using two lung ultrasound datasets and demonstrate that our model is capable of achieving performance on par with SMEs in pneumothorax identification. We then developed an iOS application that runs our full system in less than 4 seconds on an iPad Pro, and less than 8 seconds on an iPhone 13 Pro, labeling key regions in the lung sonogram to provide interpretable diagnoses.

Visual analytics systems enable highly interactive exploratory data analysis. Across a range of fields, these technologies have been successfully employed to help users learn from complex data. However, these same exploratory visualization techniques make it easy for users to discover spurious findings. This paper proposes new methods to monitor a user’s analytic focus during visual analysis of structured datasets and use it to surface relevant articles that contextualize the visualized findings. Motivated by interactive analyses of electronic health data, this paper introduces a formal model of analytic focus, a computational approach to dynamically update the focus model at the time of user interaction, and a prototype application that leverages this model to surface relevant medical publications to users during visual analysis of a large corpus of medical records. Evaluation results with 24 users show that the modeling approach has high levels of accuracy and is able to surface highly relevant medical abstracts.

Modal logics are primary formalisms for multi-agent systems but major reasoning tasks in such logics are intractable, which impedes applications of multi-agent modal logics such as automatic planning. One technique of tackling the intractability is to identify a fragment called a normal form of multiagent logics such that it is expressive but tractable for reasoning tasks such as entailment checking, bounded conjunction transformation and forgetting. For instance, DNF of propositional logic is tractable for these reasoning tasks. In this paper, we first introduce a notion of logical separability and then define a novel disjunctive normal form SDNF for the multiagent logic Kn, which overcomes some shortcomings of existing approaches. In particular, we show that every modal formula in Kn can be equivalently casted as a formula in SDNF, major reasoning tasks tractable in propositional DNF are also tractable in SDNF, and moreover, formulas in SDNF enjoy the property of logical separability. To demonstrate the usefulness of our approach, we apply SDNF in multi-agent epistemic planning. Finally, we extend these results to three more complex multi-agent logics Dn, K45n and KD45n.