A5101494707- Natural Language Processing Techniques
- Speech and dialogue systems
- Speech Recognition and Synthesis
- Advanced Computational Techniques and Applications
- Topic Modeling
- Educational Technology and Assessment
- Advanced Decision-Making Techniques
- Web Data Mining and Analysis
- Robotic Path Planning Algorithms
- Advanced Neural Network Applications
- Distributed and Parallel Computing Systems
- Mathematics, Computing, and Information Processing
- Private Equity and Venture Capital
- Recommender Systems and Techniques
- Web visibility and informetrics
- Educational Research and Pedagogy
- Collaboration in agile enterprises
- Robotics and Sensor-Based Localization
- Evaluation and Optimization Models
- Music and Audio Processing
- FinTech, Crowdfunding, Digital Finance
- Innovative Educational Techniques
- Management Theory and Practice
- Entrepreneurship Studies and Influences
- Educational Systems and Policies
Shandong Jiaotong University
2024
Shanghai Jiao Tong University
2024
University of Electronic Science and Technology of China
2022
New York University
2021
Tsinghua University
2019-2021
University of Cambridge
2019-2020
Chinese University of Hong Kong
2017
Qiqihar Medical University
2017
National Central University
2007
Shenyang University of Technology
2004
Few-shot learning aims to train models that can recognize novel classes given just a handful of labeled examples, known as the support set. While field has seen notable advances in recent years, they have often focused on multi-class image classification. Audio, contrast, is multi-label due overlapping sounds, resulting unique properties such polyphony and signal-to-noise ratios (SNR). This leads unanswered questions concerning impact audio may few-shot system design, performance,...
Existing Large Language Models (LLMs) usually remain static after deployment, which might make it hard to inject new knowledge into the model. We aim build models containing a considerable portion of self-updatable parameters, enabling model integrate effectively and efficiently. To this end, we introduce MEMORYLLM, that comprises transformer fixed-size memory pool within latent space transformer. MEMORYLLM can self-update with text memorize injected earlier. Our evaluations demonstrate...
Teaching to improve student models (e.g., knowledge distillation) is an extensively studied methodology in LLMs. However, for humans, teaching not only improves students but also teachers. We ask: Can LLMs learn by (LbT)? If yes, we can potentially unlock the possibility of continuously advancing without solely relying on human-produced data or stronger models. In this paper, provide a preliminary exploration ambitious agenda. show that LbT ideas be incorporated into existing LLM...
Abstract Learning evaluation is an effective method, which plays important role in the network education system. But most of current learning methods still use traditional university system, do not take into account web-based characteristics, and they are difficult to fit rapid development interuniversity collaborative based on network. Fuzzy comprehensive method used evaluate combination fuzzy theory analytic hierarchy process. Analytic process determine weight factors each layer carry out...
Automatic spoken language assessment (SLA) is a challenging problem due to the large variations in learner speech combined with limited resources.These issues are even more problematic when considering children learning language, higher levels of acoustic and lexical variability, code-switching compared adult data.This paper describes ALTA system for INTERSPEECH 2020 Shared Task on Speech Recognition Non-Native Children's Speech.The data this task consists examination recordings Italian...
Spatial planning in cluttered environments is crucial for mobile systems, particularly agile quadrotors. Existing methods, both optimization-based and learning-based, often focus only on success rates specific lack a unified platform with tasks of varying difficulty. To address this, we introduce FlightBench, the first comprehensive open-source benchmark 3D spatial quadrotors, comparing classical methods emerging learning-based approaches. We also develop suite task difficulty metrics...
Teacher-student learning can be applied in automatic speech recognition for model compression and domain adaptation. This trains a student to emulate the behaviour of teacher model, only is used perform recognition. Depending on application, may differ their types, complexities, input contexts, features. In previous works, it often shown that from strong allows better than an equivalent trained with reference transcriptions. However, there has not been much investigation into whether...
Recently, bidirectional recurrent network language models (bi-RNNLMs) have been shown to outperform standard, unidirectional, neural (uni-RNNLMs) on a range of speech recognition tasks. This indicates that future word context information beyond the history can be useful. However, bi-RNNLMs pose number challenges as they make use complete previous and information. impacts both training efficiency their within lattice rescoring framework. In this paper these issues are addressed by proposing...
중소벤처기업부 자료에 의하면 2014년 기준으로 우리나라 중소벤처기업은 전체 사업체수의 99.9%를 차지하고 있으며, 종사자수의 87.8%가 중소벤처기업에서 일하고 있다. 한편 중소벤처기업 CEO의 리더십 유형과 경영성과의 관계는 수많은 선행연구의 관심사였다. 최근 중소벤처기업에 대한 금융지원과 관련하여 관계금융이 주목받고 본 연구는 국내 대표적인 전문은행인 K은행과 여신거래중인 중소벤처기업을 대상으로 CEO 리더십이 관계금융을 매개변수로 하여 경영성과에 미치는 영향에 관한 실증 연구이다. 연구 결과를 요약하면 다음과 같다. 첫째, 중소벤처기업의 CEO리더십 유형 중 거래적 리더십과 변혁적 리더십은 모두 재무성과 및 비재무성과에 정(+)의 영향을 것을 확인하였다. 이는 부정적 미친다는 일부의 선행연구와 달리 리더십도 경영성과 향상에 기여함을 입증하였다. 둘째, 관계금융의 관계분석에서 관계적 요소는 부(-)의 미치지만, 금융적 미치고 있음을 두 요소에 미친다. 교환관계를...