29 January 2020 in IEEE Transactions on Geoscience and Remote Sensing
Taiwan that is located at the junction of the Eurasian Plate and the Philippine Sea Plate is one of the most active seismic zones in the world. Devastating earthquakes have occurred around the island and have caused severe damages from time to time. To avoid the severe loss, earthquake early warning (EEW) is of great importance, and one of the most critical issues of EEW is fast and reliable detection for the presence of earthquakes. Traditional methods for earthquake detection usually use criterion-based algorithms to detect the onset of the earthquake waves. Currently, the thresholds for those criteria are usually decided empirically and may result in excessive false alarms. Obviously, false alarms can cause undue panics and diminish the credibility of the system. In this article, the recurrent neural network (RNN) models are adopted to develop a real-time EEW system. The developed system is designed to identify the occurrence of an earthquake event, and the duration of the P-wave and the S-wave. It was trained and tested using the seismograms recorded in Taiwan from 2016 to 2017. From the simulation results, the proposed scheme outperforms the traditional criterion-based schemes in terms of detection accuracy and processing time.
11 April 2022 in IEEE/ACM Transactions on Audio, Speech, and Language Processing
Transformer-based models have led to significant innovation in various classic and practical subjects, including speech processing, natural language processing, and computer vision. On top of the Transformer, attention-based end-to-end automatic speech recognition (ASR) models have become a popular fashion in recent years. Specifically, an emergent research topic is non-autoregressive modeling, which can achieve fast inference speed and obtain competitive performance when compared with conventional autoregressive methods. In addition, in the context of natural language processing, the bidirectional encoder representations from Transformers (BERT) model and its variants have received widespread attention, partially due to their ability to infer contextualized word representations and obtain superior performances of downstream tasks through simple fine-tuning. However, to our knowledge, leveraging the synergistic power of non-autoregressive modeling and pre-trained language model for ASR remains relatively underexplored. In this regard, this study presents a novel pre-trained language model-based non-autoregressive ASR framework. A series of experiments were conducted on two publicly available Chinese datasets, AISHELL-1 and AISHELL-2, to demonstrate competitive or superior results of the proposed ASR models when compared with well-practiced baseline systems. In addition, a set of comparative experiments is likewise carried out with different settings to analyze the performance of the proposed framework.
19 October 2017 in IEEE/ACM Transactions on Audio, Speech, and Language Processing
In the context of natural language processing, representation learning has emerged as a newly active research subject because of its excellent performance in many applications. Learning representations of words is a pioneering study in this school of research. However, paragraph (or sentence and document) embedding learning is more suitable/reasonable for some realistic tasks such as document summarization. Nevertheless, classic paragraph embedding methods infer the representation of a given paragraph by considering all of the words occurring in the paragraph. Consequently, those stop or function words that occur frequently may mislead the embedding learning process to produce a misty paragraph representation. Motivated by these observations, our major contributions in this paper are threefold. First, we propose a novel unsupervised paragraph embedding method, named the essence vector (EV) model, which aims at not only distilling the most representative information from a paragraph but also excluding the general background information to produce a more informative low-dimensional vector representation for the paragraph of interest. Second, in view of the increasing importance of spoken content processing, an extension of the EV model, named the denoising essence vector (D-EV) model, is proposed. The D-EV model not only inherits the advantages of the EV model but also can infer a more robust representation for a given spoken paragraph against imperfect speech recognition. Third, a new summarization framework, which can take both relevance and redundancy information into account simultaneously, is also introduced. We evaluate the proposed embedding methods (i.e., EV and D-EV) and the summarization framework on two benchmark summarization corpora. The experimental results demonstrate the effectiveness and applicability of the proposed framework in relation to several well-practiced and state-of-the-art summarization methods.
16 September 2022 in IEEE Access
In recent years, multilingual question answering has been an emergent research topic and has attracted much attention. Although systems for English and other rich-resource languages that rely on various advanced deep learning-based techniques have been highly developed, most of them in low-resource languages are impractical due to data insufficiency. Accordingly, many studies have attempted to improve the performance of low-resource languages in a zero-shot or few-shot manner based on multilingual bidirectional encoder representations from transformers (mBERT) by transferring knowledge learned from rich-resource languages to low-resource languages. Most methods require either a large amount of unlabeled data or a small set of labeled data for low-resource languages. In Wikipedia, 169 languages have less than 10,000 articles, and 48 languages have less than 1,000 articles. This reason motivates us to conduct a zero-shot multilingual question answering task under a zero-resource scenario. Thus, this study proposes a framework to fine-tune the original mBERT using data from rich-resource languages, and the resulting model can be used for low-resource languages in a zero-shot and zero-resource manner. Compared to several baseline systems, which require millions of unlabeled data for low-resource languages, the performance of our proposed framework is not only highly comparative but is also better for languages used in training.
17 August 2021 in IEEE Transactions on Geoscience and Remote Sensing
The accuracy of earthquake localization is of great importance for earthquake monitoring systems. Traditionally, numerical optimization methods are used to estimate the hypocenter location and the origin time of an earthquake in an iterative manner. The traditional methods usually depend on certain theoretical models, but the geological conditions in practice can be quite different from the presumed models. In this study, an attention-based hypocenter estimation (AHE) model was proposed to locate the hypocenter and origin time of the earthquake. Rather than using the raw waveforms, the phase picking times and the positions of the triggered stations are used as the input. The attention mechanism is adapted to reveal the correlations among the input sequence. An experimental model was trained using data collected from earthquakes in Taiwan in 2016 and 2017 and tested on data in 2018. From the results, AHE is capable of locating the hypocenter with a high degree of accuracy in terms of the distance, depth, and origin time of earthquakes.
3 December 2024 in Seismological Research Letters
In recent years, numerous studies have employed deep learning in seismology, with data-driven neural network models increasingly becoming the norm in emerging research paradigms. Alongside advancements in model architectures, several benchmark datasets have also been introduced. However, many of these datasets suffer from imbalanced distributions, particularly a scarcity of large-magnitude events, which can impair the performance of deep learning-based models. Taiwan, situated in a seismically active region experiences a high frequency of earthquakes annually. The Central Weather Administration (CWA) in Taiwan maintains comprehensive records of these seismic events. This study introduces a benchmark dataset, named CWA, specifically compiled from this extensive database. The CWA benchmark includes 331 events with magnitudes greater than five, collected from a high-density seismic network spanning from 2011 to 2021, making it well-suited for training deep learning models. In addition, the CWA benchmark features over 40 attributes and ∼500,000 seismograms, providing valuable data labels for various seismology-related tasks.