A5103264503- Topic Modeling
- Natural Language Processing Techniques
- Speech and dialogue systems
- Distributed and Parallel Computing Systems
- Advanced Malware Detection Techniques
- Reinforcement Learning in Robotics
- Cerebrovascular and Carotid Artery Diseases
- Parallel Computing and Optimization Techniques
- Multimodal Machine Learning Applications
- Scientific Computing and Data Management
- Machine Learning and Data Classification
- Anomaly Detection Techniques and Applications
- Acute Ischemic Stroke Management
- Network Security and Intrusion Detection
- Genomics and Phylogenetic Studies
Toronto Metropolitan University
2023-2024
Boston University
2003-2021
Meta (Israel)
2021
Harvard University
2003
An effective recipe for building seq2seq, non-autoregressive, task-oriented parsers to map utterances semantic frames proceeds in three steps: encoding an utterance x, predicting a frame’s length |y|, and decoding |y|-sized frame with ontology tokens. Though empirically strong, these models are typically bottlenecked by prediction, as even small inaccuracies change the syntactic characteristics of resulting frames. In our work, we propose span pointer networks, non-autoregressive which shift...
Task-oriented semantic parsing models typically have high resource requirements: to support new ontologies (i.e., intents and slots), practitioners crowdsource thousands of samples for supervised fine-tuning. Partly, this is due the structure de facto copy-generate parsers; these treat ontology labels as discrete entities, relying on parallel data extrinsically derive their meaning. In our work, we instead exploit what intrinsically know about labels; example, fact that SL:TIME_ZONE has...
Spoken Language Understanding (SLU) is a critical component of voice assistants; it consists converting speech to semantic parses for task execution. Previous works have explored end-to-end models improve the quality and robustness SLU with Deliberation, however these remained autoregressive, resulting in higher latencies. In this work we introduce PRoDeliberation, novel method leveraging Connectionist Temporal Classification-based decoding strategy as well denoising objective train robust...
In the realm of social network analysis and mining, recommendation systems have become indispensable algorithms in assisting users industries navigating available contents or products various domains getting most personalized recommendations to their interests preferences. However, if input data has been generated by malicious users, that poses a significant challenge recommender systems' reliability efficiency. One main threats is shilling attacks. Shilling attacks tend manipulate poison...
Spoken language understanding (SLU) is a important field between the Speech and NLP community focused on converting users' speech utterance into an executable semantic parse. In order to facilitate open research in this space, we introduce 1st Language Understanding challenge hosted at ICASSP 2023. We leverage newly released SLU dataset STOP [1]. challenge, participants are asked compete 3 tracks of relevant (1) Quality: build highest performance model (2) On-device: quality under 15M...
The Research Computing Services (RCS) group at Boston University (BU), developed a benchmark suite to evaluate the performance of newer hardware under consideration for purchase BU Shared Cluster (SCC). custom benchmarks is used generate metrics that are representative highly diverse types jobs run on cluster. results make informed decisions about upgrades in order provide best balance and value cluster users. In this paper we discuss present reasons creating suite, general architecture...