- Natural Language Processing Techniques
- Topic Modeling
- Multimodal Machine Learning Applications
- Speech and dialogue systems
- Radiology practices and education
- COVID-19 diagnosis using AI
- Lung Cancer Diagnosis and Treatment
- Semantic Web and Ontologies
- Hate Speech and Cyberbullying Detection
- Adversarial Robustness in Machine Learning
- Software Engineering Research
- Explainable Artificial Intelligence (XAI)
- Advanced Graph Neural Networks
Google (United States)
2019-2023
California University of Pennsylvania
2018
BackgroundDeep learning has the potential to augment use of chest radiography in clinical radiology, but challenges include poor generalizability, spectrum bias, and difficulty comparing across studies.PurposeTo develop evaluate deep models for radiograph interpretation by using radiologist-adjudicated reference standards.Materials MethodsDeep were developed detect four findings (pneumothorax, opacity, nodule or mass, fracture) on frontal radiographs. This retrospective study used two data...
In this report, we present the latest model of Gemini family, 1.5 Pro, a highly compute-efficient multimodal mixture-of-experts capable recalling and reasoning over fine-grained information from millions tokens context, including multiple long documents hours video audio. Pro achieves near-perfect recall on long-context retrieval tasks across modalities, improves state-of-the-art in long-document QA, long-video QA ASR, matches or surpasses 1.0 Ultra's performance broad set benchmarks....
Language models (LMs) pretrained on large corpora of text from the web have been observed to contain amounts various types knowledge about world. This observation has led a new and exciting paradigm in graph construction where, instead manual curation or mining, one extracts parameters an LM. Recently, it shown that finetuning LMs set factual makes them produce better answers queries different set, thus making finetuned good candidate for extraction and, consequently, construction. In this...
Words are polysemous and multi-faceted, with many shades of meanings. We suggest that sparse distributed representations more suitable than other, commonly used, (dense) to express these multiple facets, present Category Builder, a working system that, as we show, makes use support multi-faceted lexical representations. argue the set expansion task is well suited study meaning distinctions since word may belong sets different reason for membership in each. therefore exhibit performance...
Foundation models, i.e. large neural networks pre-trained on text corpora, have revolutionized NLP. They can be instructed directly (e.g. (arXiv:2005.14165)) - this is called hard prompting and they tuned using very little data (arXiv:2104.08691)) technique soft prompting. We seek to leverage their capabilities detect policy violations. Our contributions are: identify a prompt that adapts chain-of-thought violation tasks. This produces classifications, along with extractive explanations...
The prevalence and impact of toxic discussions online have made content moderation crucial.Automated systems can play a vital role in identifying toxicity, reducing the reliance on human moderation.Nevertheless, comments for diverse communities continues to present challenges that are addressed this paper.The two-part goal study is to(1)identify intuitive variances from annotator disagreement using quantitative analysis (2)model subjectivity these viewpoints.To achieve our goal, we published...
Words are polysemous and multi-faceted, with many shades of meanings. We suggest that sparse distributed representations more suitable than other, commonly used, (dense) to express these multiple facets, present Category Builder, a working system that, as we show, makes use support multi-faceted lexical representations. argue the set expansion task is well suited study meaning distinctions since word may belong sets different reason for membership in each. therefore exhibit performance...