A5083688498- Topic Modeling
- Speech and dialogue systems
- Natural Language Processing Techniques
- Misinformation and Its Impacts
- Sentiment Analysis and Opinion Mining
- Social Media and Politics
- Advanced Text Analysis Techniques
- Opinion Dynamics and Social Influence
- Language, Metaphor, and Cognition
- Multi-Agent Systems and Negotiation
- Software Engineering Research
- Advanced Malware Detection Techniques
- Complex Network Analysis Techniques
- Digital Communication and Language
- Language, Discourse, Communication Strategies
- Spam and Phishing Detection
- Data Visualization and Analytics
- Digital Marketing and Social Media
- Mental Health via Writing
- AI in Service Interactions
- Hate Speech and Cyberbullying Detection
- Semantic Web and Ontologies
- Computational and Text Analysis Methods
- Discourse Analysis in Language Studies
- Explainable Artificial Intelligence (XAI)
University of North Carolina at Charlotte
2017-2024
IT University of Copenhagen
2023
Tokyo Institute of Technology
2023
Administration for Community Living
2023
American Jewish Committee
2023
University at Albany, State University of New York
2009-2021
Albany State University
2010-2021
University of Washington
2021
University of Kentucky
2021
Jordan University of Science and Technology
2019
Sebastian Gehrmann, Tosin Adewumi, Karmanya Aggarwal, Pawan Sasanka Ammanamanchi, Anuoluwapo Aremu, Antoine Bosselut, Khyathi Raghavi Chandu, Miruna-Adriana Clinciu, Dipanjan Das, Kaustubh Dhole, Wanyu Du, Esin Durmus, Ondřej Dušek, Chris Chinenye Emezue, Varun Gangal, Cristina Garbacea, Tatsunori Hashimoto, Yufang Hou, Yacine Jernite, Harsh Jhamtani, Yangfeng Ji, Shailza Jolly, Mihir Kale, Dhruv Kumar, Faisal Ladhak, Aman Madaan, Mounica Maddela, Khyati Mahajan, Saad Mahamood, Bodhisattwa...
Anchoring effect is the tendency to focus too heavily on one piece of information when making decisions. In this paper, we present a novel, systematic study and resulting analyses that investigate effects anchoring human decision-making using visual analytic systems. Visual analytics interfaces typically contain multiple views various aspects such as spatial, temporal, categorical. These are designed complex, heterogeneous data in accessible forms aid decision-making. However, often hindered...
We introduce GEM, a living benchmark for natural language Generation (NLG), its Evaluation, and Metrics. Measuring progress in NLG relies on constantly evolving ecosystem of automated metrics, datasets, human evaluation standards. Due to this moving target, new models often still evaluate divergent anglo-centric corpora with well-established, but flawed, metrics. This disconnect makes it challenging identify the limitations current opportunities progress. Addressing limitation, GEM provides...
Background. Engagement has been identified as a crucial component of learning in games research. However, the conceptualization and operationalization engagement vary widely literature. Many valuable approaches illuminate ways which presence, flow, arousal, participation, other concepts constitute or contribute to engagement. few studies examine multiple conceptualizations same project. Method. This article discusses results two experiments that measure five different ways: survey...
Task 1 in the International Workshop SemEval 2018, Affect Tweets, introduces five subtasks (El-reg, El-oc, V-reg, V-oc, and E-c) to detect intensity of emotions English, Arabic, Spanish tweets. This paper describes TeamUNCC's system English Arabic Our approach is novel that we present same architecture for all both Arabic. The main input a combination word2vec doc2vec embeddings set psycholinguistic features (e.g. from AffectTweets Weka-package). We apply fully connected neural network...
This is a report on the NSF Future Directions Workshop Automatic Evaluation of Dialog. The workshop explored current state art along with its limitations and suggested promising directions for future work in this important very rapidly changing area research.
The increasing reliance on multimedia content has heightened the need for inclusive technologies, particularly individuals with hearing impairments. This paper presents a Real-Time Caption Generator that leverages Whisper ASR model to convert video audio into accurate, synchronized captions <500ms end-to-end latency. system addresses accessibility gaps through React-based interactive interface, enabling users customize caption display settings (font style, size, and opacity) optimal...
Social media have been increasingly adopted by health agencies to disseminate information, interact with the public, and understand public opinion. Among them, Centers for Disease Control Prevention (CDC) is one of first US government adopt social during emergencies crisis. It had active on Twitter 2016 Zika epidemic that caused 5168 domestic noncongenital cases in United States.The aim this study was quantify temporal variabilities CDC's tweeting activities throughout epidemic, engagement...
Emotion detection is one of the most challenging problems in automated understand language. Understanding human emotions using text without facial expression considered a complicated task. Therefore, building machine that understands context sentences and differentiates between has motivated learning community recently. We propose system to detect deep approaches. The main input combination GloVe word embeddings, BERT Embeddings set psycholinguistic features (e.g. from AffectiveTweets...
One of the hardest problems in area Natural Language Processing and Artificial Intelligence is automatically generating language that coherent understandable to humans. Teaching machines how converse as humans do falls under broad umbrella Generation. Recent years have seen unprecedented growth number research articles published on this subject conferences journals both by academic industry researchers. There also been several workshops organized alongside top-tier NLP dedicated specifically...
We describe a novel study of decision-making processes around misinformation on social media. Using custom-built visual analytic system, we presented users with news content from media accounts variety outlets, including outlets engaged in distributing misinformation. conducted controlled experiments to regarding the veracity these and tested role confirmation bias (the tendency ignore contradicting information) uncertainty information human processes. Our findings reveal that presence...
To overcome the limitations of automated metrics (e.g. BLEU, METEOR) for evaluating dialogue systems, researchers typically use human judgments to provide convergent evidence. While it has been demonstrated that can suffer from inconsistency ratings, extant research also found design evaluation task affects consistency and quality judgments. We conduct a between-subjects study understand impact four experiment conditions on ratings system output. In addition discrete continuous scale we with...
We present Verifi2, a visual analytic system to support the investigation of misinformation on social media. Various models and studies have emerged from multiple disciplines detect or understand effects misinformation. However, there is still lack intuitive accessible tools that help media users distinguish verified news. Verifi2 uses state-of-the-art computational methods highlight linguistic, network, image features can suspicious news accounts. By exploring source document level in...
Writing software exploits is an important practice for offensive security analysts to investigate and prevent attacks. In particular, shellcodes are especially time-consuming a technical challenge, as they written in assembly language. this work, we address the task of automatically generating shellcodes, starting purely from descriptions natural language, by proposing approach based on Neural Machine Translation (NMT). We then present empirical study using novel dataset (Shellcode_IA32),...
Abstract In this paper, we describe a novel approach to computational modeling and understanding of social cultural phenomena in multi-party dialogues. We developed two-tier which first detect classify certain sociolinguistic behaviors, including topic control, disagreement, involvement, that serve as first-order models from presence the higher level roles, such leadership, may be inferred.
The internet and the high use of social media have enabled modern-day journalism to publish, share spread news that is difficult distinguish if it true or fake. Defining “fake news” not well established yet, however, can be categorized under several labels: false, biased, framed mislead readers are characterized as propaganda. Digital content production technologies with logical fallacies emotional language used propaganda techniques gain more audience. Recently, researchers proposed deep...
Humans quite frequently interact with conversational agents. The rapid advancement in generative language modeling through neural networks has helped advance the creation of intelligent Researchers typically evaluate output their models crowdsourced judgments, but there are no established best practices for conducting such studies. Moreover, it is unclear if cognitive biases decision-making affecting workers' judgments when they undertake these tasks. To investigate, we conducted a...
Writing exploits for security assessment is a challenging task. The writer needs to master programming and obfuscation techniques develop successful exploit. To make the task easier, we propose an approach (EVIL) automatically generate in assembly/Python language from descriptions natural language. leverages Neural Machine Translation (NMT) dataset that developed this work. We present extensive experimental study evaluate feasibility of EVIL, using both automatic manual analysis, at...