A5014763831- Functional Brain Connectivity Studies
- Neural dynamics and brain function
- Natural Language Processing Techniques
- Topic Modeling
- EEG and Brain-Computer Interfaces
- Advanced MRI Techniques and Applications
- Neural and Behavioral Psychology Studies
- Text and Document Classification Technologies
- Data Quality and Management
- Visual perception and processing mechanisms
- Advanced Neuroimaging Techniques and Applications
- Memory and Neural Mechanisms
- Heart Rate Variability and Autonomic Control
- Face Recognition and Perception
- Atomic and Subatomic Physics Research
- Photoreceptor and optogenetics research
- Text Readability and Simplification
- Speech and dialogue systems
- Traumatic Brain Injury and Neurovascular Disturbances
- Sleep and related disorders
- Multimodal Machine Learning Applications
- Mental Health Research Topics
- Muscle activation and electromyography studies
- Vaccine Coverage and Hesitancy
- Optical Imaging and Spectroscopy Techniques
Aston University
2024-2025
Duolingo (United States)
2021-2023
University of Birmingham
2013-2022
Finnish Medical Society Duodecim
2021
University of Pennsylvania
2018-2021
University of Illinois Urbana-Champaign
2014-2020
California University of Pennsylvania
2018-2020
Coalition for Epidemic Preparedness Innovations
2020
Indian Institute of Technology Kanpur
2020
University of Pittsburgh
2020
Recent work has exhibited the surprising cross-lingual abilities of multilingual BERT (M-BERT) -- since it is trained without any objective and with no aligned data. In this work, we provide a comprehensive study contribution different components in M-BERT to its ability. We impact linguistic properties languages, architecture model, learning objectives. The experimental done context three typologically languages Spanish, Hindi, Russian using two conceptually NLP tasks, textual entailment...
Massed synchronised neuronal firing is detrimental to information processing. When networks of task-irrelevant neurons fire in unison, they mask the signal generated by task-critical neurons. On a macroscopic level, such synchronisation can contribute alpha/beta (8-30 Hz) oscillations. Reducing amplitude these oscillations, therefore, may enhance Here, we test this hypothesis. Twenty-one participants completed an associative memory task while undergoing simultaneous EEG-fMRI recordings....
Abstract We take a step towards addressing the under- representation of African continent in NLP research by bringing together different stakeholders to create first large, publicly available, high-quality dataset for named entity recognition (NER) ten languages. detail characteristics these languages help researchers and practitioners better understand challenges they pose NER tasks. analyze our datasets conduct an extensive empirical evaluation state- of-the-art methods across both...
Abstract Purpose To estimate the importance of respiratory and cardiac effects on signal variability found in functional magnetic resonance imaging data recorded from brainstem. Materials Methods A modified version retrospective image correction (RETROICOR) method (Glover et al, [2000] Magn Reson Med 44:162–167) was implemented resting brainstem echo‐planar (EPI) 12 subjects. Fourier series were fitted to based recordings (pulseoximetry turbine), including multiplicative terms that accounted...
Recent work in NLP has attempted to deal with low-resource languages but still assumed a resource level that is not present for most languages, e.g., the availability of Wikipedia target language. We propose simple method cross-lingual named entity recognition (NER) works well settings very minimal resources. Our approach makes use lexicon "translate" annotated data available one or several high language(s) into language, and learns standard monolingual NER model there. Further, when our can...
Unambiguous interpretation of changes in the BOLD signal is challenging because complex neurovascular coupling that translates neuronal activity into subsequent haemodynamic response. In particular, neurophysiological origin negative response (NBR) remains incompletely understood. Here, we simultaneously recorded BOLD, EEG and cerebral blood flow (CBF) responses to 10 s blocks unilateral median nerve stimulation (MNS) order interrogate NBR. Both CBF MNS were observed same region ipsilateral...
The human brain is continually, dynamically active and spontaneous fluctuations in this activity play a functional role affecting both behavioural neuronal responses. However, the mechanisms through which occurs remain poorly understood. Simultaneous EEG-fMRI promising technique to study how modulates brain's response stimulation, as temporal indices of ongoing cortical excitability can be integrated with spatially localised evoked Here we demonstrate an interaction between power...
Named Entity Recognition (NER) models for language L are typically trained using annotated data in that language.We study cross-lingual NER, where a model NER is on another, source, (or multiple source languages).We introduce independent method building wikification, technique grounds words and phrases non-English text into English Wikipedia entries.Thus, mentions any can be described set of categories FreeBase types, yielding, as we show, strong language-independent features.With this...
Multilingual BERT (M-BERT) has been a huge success in both supervised and zero-shot cross-lingual transfer learning. However, this is focused only on the top 104 languages Wikipedia it was trained on. In paper, we propose simple but effective approach to extend M-BERT E-MBERT so can benefit any new language, show that our aids are already as well. We perform an extensive set of experiments with Named Entity Recognition (NER) 27 languages, 16 which M-BERT, average increase about 6% F1 23%...
Investigations into the blood oxygenation level-dependent (BOLD) functional MRI signal have used respiratory challenges with aim of probing cerebrovascular physiology. Such altered inspired partial pressures either carbon dioxide or oxygen, typically to a fixed and constant level (fixed challenge (FIC)). The resulting end-tidal gas then depend on subject's metabolism ventilatory responses. In contrast, dynamic forcing (DEF) rapidly independently sets oxygen desired levels by altering...
fMRI is the foremost technique for noninvasive measurement of human brain function. However, its utility limited by an incomplete understanding relationship between neuronal activity and hemodynamic response. Though primary peak response modulated activity, origin typically negative poststimulus signal poorly understood amplitude assumed to covary with We use simultaneous recordings EEG blood oxygenation level-dependent (BOLD) cerebral flow (CBF) during unilateral median nerve stimulation...
Directing attention helps to extract relevant information and suppress distracters. Alpha brain oscillations (8-12 Hz) are crucial for this process, with power decreases facilitating processing of important increases inhibiting regions irrelevant information. Evidence phenomenon arises from visual studies (Worden et al., 2000); however, the effect also exists in other modalities, including somatosensory system (Haegens 2011) intersensory tasks (Foxe Snyder, 2011). We investigated human...
Negative BOLD fMRI responses (NBR) occur commonly in sensory cortex and default mode network regions but remain poorly utilized as a marker of brain function due to an incomplete understanding. To better understand how NBR manifest across the brain, compare between different stimuli they are modulated by changes task demand, we recorded during trials visual, auditory, or somatosensory stimulation, delivered either alone concurrent pairs. Twenty young-adult participants were cued attend...
The stimulus-evoked response is the principle measure used to elucidate timing and spatial location of human brain activity. Brain behavioural responses pain are influenced by multiple intrinsic extrinsic factors display considerable, natural trial-by-trial variability. However, because neuronal sources this variability poorly understood functional information it contains under-exploited for understanding relationship between function behaviour. We recorded simultaneous EEG-fMRI during rest...
Self-imposed short sleep durations are increasingly commonplace in society, and have considerable health performance implications for individuals. Reduced duration over multiple nights has similar behavioural effects to those observed following acute total deprivation, suggesting that lack of affects brain function cumulatively. A link between habitual patterns functional connectivity previously been observed, the effect on brain's intrinsic architecture may provide a status cognition....
Post-stimulus undershoots, negative responses following cessation of stimulation, are widely observed in functional magnetic resonance (fMRI) blood oxygenation level dependent (BOLD) data. However, the debate surrounding whether origin this response phase is neuronal or vascular, and it provides functionally relevant information, that additional to what contained primary response, means undershoots overlooked. We simultaneously recorded electroencephalography (EEG), BOLD cerebral blood-flow...
Supervised machine learning assumes the availability of fully-labeled data, but in many cases, such as low-resource languages, only data available is partially annotated. We study problem Named Entity Recognition (NER) with annotated training which a fraction named entities are labeled, and all other tokens, or otherwise, labeled non-entity by default. In order to train on this noisy dataset, we need distinguish between true false negatives. To end, introduce constraint-driven iterative...