A5047997552- Natural Language Processing Techniques
- Topic Modeling
- Speech Recognition and Synthesis
- Neural dynamics and brain function
- Service-Oriented Architecture and Web Services
- Fault Detection and Control Systems
- UAV Applications and Optimization
- Advanced Chemical Sensor Technologies
- Distributed Control Multi-Agent Systems
- stochastic dynamics and bifurcation
- Advanced Memory and Neural Computing
- Barrier Structure and Function Studies
- Gas Sensing Nanomaterials and Sensors
- Advanced Sensor and Control Systems
- Refrigeration and Air Conditioning Technologies
- Cloud Computing and Resource Management
- Advanced Computational Techniques and Applications
- Music and Audio Processing
- Algorithms and Data Compression
- Acoustic Wave Resonator Technologies
- Thermodynamic and Exergetic Analyses of Power and Cooling Systems
- Guidance and Control Systems
- Urban Stormwater Management Solutions
- Flood Risk Assessment and Management
- Analog and Mixed-Signal Circuit Design
China Shipbuilding Industry Corporation (China)
2024
Harbin Institute of Technology
2011-2024
Jiangnan University
2024
Bryan College
2023
Texas A&M University
2014-2023
PLA Army Engineering University
2020-2023
Zhengzhou University
2021-2023
China Jiliang University
2023
Jilin University
2008-2022
Beijing University of Posts and Telecommunications
2006-2022
Microvascular hyperpermeability that occurs at the level of blood-brain barrier (BBB) often leads to vasogenic brain edema and elevated intracranial pressure following traumatic injury (TBI). At a cellular level, tight junction proteins (TJPs) between neighboring endothelial cells maintain integrity BBB via TJ associated particularly, zonula occludens-1 (ZO-1) binds transmembrane TJPs actin cytoskeleton intracellularly. The pro-inflammatory cytokine, interleukin-1β (IL-1β) as well...
We introduce a novel precedence reordering approach based on dependency parser to statistical machine translation systems.Similar other preprocessing approaches, our method can efficiently incorporate linguistic knowledge into SMT systems without increasing the complexity of decoding.For set five subject-object-verb (SOV) order languages, we show significant improvements in BLEU scores when translating from English, compared state-of-the-art phrase-based systems.
The yellow boxfish ( Ostracion cubicus) exhibits a combination of derived morphological traits specialized for coral reef environments and ancestral characteristics, including fused dermal plate. Contradictory evolutionary evidence hinders true classification O. cubicus. To clarify its position within Tetraodontiformes, chromosome-level genome assembly was generated, representing the most contiguous complete to date this lineage. Notably, cubicus possessed largest order primarily due...
The paper presents an in-depth analysis of a less known interaction between Kneser-Ney smoothing and entropy pruning that leads to severe degradation in language model performance under aggressive regimes. Experiments data-rich setup such as google.com voice search show significant impact WER well: Katz models 0.1% their original impacts speech recognition accuracy significantly, approx. 10% relative.
Abstract During the 2007-2008 global financial crisis, many countries enacted clean energy policies as a part of their economic stimulus packages. These are believed to have contributed significant reduction in CO 2 intensity electricity. Here we conduct retrospective overview and evaluation policies’ effectiveness reducing We utilize governance capacity measure policy implementation stringency, interaction between number categorized adjust variables for effectiveness. distinguish short-...
We study the impact of richer syntactic dependencies on performance structured language model (SLM) along three dimensions: parsing accuracy (LP/LR), perplexity (PPL) and word-error-rate (WER, N-best re-scoring). show that our models achieve an improvement in LP/LR, PPL and/or WER over reported baseline results using SLM UPenn Treebank Wall Street Journal (WSJ) corpora, respectively. Analysis shows correlation between quality parser (as measured by precision/recall) (PPL WER). A remarkable...
The paper presents an empirical exploration of google.com query stream language modeling. We describe the normalization typed resulting in out-of-vocabulary (OoV) rates below 1% for a one million word vocabulary. present comprehensive set experiments that guided design decisions voice search service. In process we re-discovered less known interaction between Kneser-Ney smoothing and entropy pruning, found evidence hints at non-stationarity stream, as well strong dependence on various English...
A novel blocks placement strategy is developed for solving some shortages in recent Hadoop. With the Strategy, optimal Data nodes, according to real-time situation of are chosen dynamically achieve load balancing. Finally, simulation experiments indicate that our behaves much better than HDFS when coherent time.
Compared with traditional optical devices, metasurfaces have attracted extensive attention due to their unique electromagnetic properties as well advantages of thinness, ease integration, and low loss. However, structural modeling, simulation calculations, parameter optimization processes are often required for metasurface design by methods, which consume time computing resources. Here, we propose an inverse method based on deep tandem neural networks speed up the process metasurfaces. This...
This paper proposes a new discriminative training method, called minimum sample risk (MSR), of estimating parameters language models for text input. While most existing methods use loss function that can be optimized easily but approaches only approximately to the objective error rate, MSR minimizes directly using heuristic procedure. Evaluations on task Japanese input show handle large number features and samples; it significantly outperforms regular trigram model trained maximum likelihood...
This paper investigates semi-supervised methods for discriminative language modeling, whereby n-best lists are "hallucinated" given reference text and then used training n-gram models using the perceptron algorithm. We perform controlled experiments on a very strong baseline English CTS system, comparing three simulating ASR output, compare results with "real" list output from recognizer. find that based extracting phrasal cohorts - similar to machine translation phrase tables yielded...
Drift compensation is an important issue for metal oxide semiconductor (MOS) gas sensor arrays. General machine learning methods require constant calibration and a large amount of label data. At the same time, recalibration will cause lot costs, difficult to obtain in practice. In this paper, novel drift method based on balanced distribution adaptation (BDA) proposed. First, BDA can adjust conditional marginal between two domains through weight balance factor, thereby more effectively...
We present our work on semi-supervised learning of discriminative language models where the negative examples for sentences in a text corpus are generated using confusion Turkish at various granularities, specifically, word, sub-word, syllable and phone levels. experiment with different sampling strategies to select competing hypotheses training variant perceptron algorithm. find that morph-based sample selection strategy aiming match error distribution baseline ASR system gives best...
The flammable and explosive property of hydrogen is the main danger in its safe use, storage transportation. In this paper, a novel monitoring system designed based on principle semiconductor, catalytic combustion heat-conducting gas sensors. Also, sensor will inevitably fail due to nature sensitive materials long-time process. To ensure accuracy reliability concentration measurement, fault diagnosis reconfiguration strategy for array moving window component analysis extreme learning machine...