对文本进行分词和词嵌入通常是中文命名实体识别的第一步,但中文的词与词之间没有明确的分界符,专业词及生僻词等未收录词(Out of Vocabulary,OOV)严重干扰了词向量的计算,基于词向量嵌入的模型性能极易受到分词效果的影响。同时现有模...对文本进行分词和词嵌入通常是中文命名实体识别的第一步,但中文的词与词之间没有明确的分界符,专业词及生僻词等未收录词(Out of Vocabulary,OOV)严重干扰了词向量的计算,基于词向量嵌入的模型性能极易受到分词效果的影响。同时现有模型大多使用循环神经网络,计算速度较慢,很难达到工业应用的要求。针对上述问题,构建了一个基于注意力机制和卷积神经网络的命名实体识别模型,即LAC-DGLU。针对分词依赖的问题,提出了一种基于局部注意力卷积(Local Attention Convolution,LAC)的字嵌入算法,减轻了模型对分词效果的依赖。针对计算速度较慢的问题,使用了一种带门结构的卷积神经网络,即膨胀门控线性单元(Dilated Gated Linear Unit,DGLU),提高了模型的计算速度。在多个数据集上的实验结果显示,该模型相比现有最优模型F1值提高了0.2%~2%,训练速度可以达到现有最优模型的1.4~1.9倍。展开更多
In order to resolve the conflict between the limited resources of embedded devices and the growing amount of massive image data to be shown, a solution for fast images rendering in embedded devices is proposed and imp...In order to resolve the conflict between the limited resources of embedded devices and the growing amount of massive image data to be shown, a solution for fast images rendering in embedded devices is proposed and implemented. First, an improved algorithm of a multi-resolution file-pyramid construction which is used for the organization of massive image data is presented. Then, a strategy, adopting technologies such as view-dependent levels of detail, target-tiles quick search and tiles seamless connection, is presented for fast scheduling and viewing of images. The results show that compared with the solution of multi-scale image representations based on wavelet, the proposed solution can improve the rendering speed, and the rendering speed does not depend on the image size, though it increases some data storage space. And the proposed solution is suitable for embedded devices and friendly experience.展开更多
Optical fiber acceleration seismometer as an important instrument can offer high sensitivity, anti-jamming and non-touched advantage which has an extensive application field. Its signal processing ability will decide ...Optical fiber acceleration seismometer as an important instrument can offer high sensitivity, anti-jamming and non-touched advantage which has an extensive application field. Its signal processing ability will decide whole system’s performance to some extent because it will affect directly the factors such as resolving power, precision and dynamic range. The signal processing is usually realized by analog circuits which was more inferior in stability, flexibility and anti-jamming to digital processing system. A digital processing system of optical fiber acceleration seismometer has been designed based on the embedded system design scheme. Synthetic-heterodyne demodulation has been studied, and signal processing has been realized. The double processors of ARM and DSP are employed to implement respectively the system control and signal processing, and to provide the output interfaces such as LCD, DAC and Ethernet interface. This system can vary with the measured signal in real time and linearly, and its work frequency bandwidth is between 10Hz and 1kHz. The system has better anti-jamming ability and can work normally when the SNR is 40dB.展开更多
Sign Writing, a writing system for sign language, is becoming a useful and convenient communication aid for people who are deaf. Principally, people who are deaf find it difficult to communicate with the hearing commu...Sign Writing, a writing system for sign language, is becoming a useful and convenient communication aid for people who are deaf. Principally, people who are deaf find it difficult to communicate with the hearing community and due to recent technological advancement they communicate amongst themselves and with the hearing community via text messaging on mobile phones. Existing messaging function is limited to writing based on the Roman alphabets or pictographic languages like Mandarin; writing in signs is deemed to be deficient in a mobile context. Hence, the aim of this paper is to examine the feasibility of writing and reading text messages in signs as an alternative communication mean besides Short Messaging Service （SMS）. Initial experimental results have significantly exemplified that sign writing gains well acceptance and is preferred among the hearing-impaired community to communicate within or between the hearing communities.展开更多
AIMTo assess KRAS G12D mutation detection by droplet digital PCR (ddPCR) in stool-derived DNA from colorectal cancer (CRC) patients.METHODSIn this study, tumor tissue and stool samples were collected from 70 patients ...AIMTo assess KRAS G12D mutation detection by droplet digital PCR (ddPCR) in stool-derived DNA from colorectal cancer (CRC) patients.METHODSIn this study, tumor tissue and stool samples were collected from 70 patients with stage I-IV CRC diagnosed by preoperative biopsy. KRAS mutational status was determined by pyrosequencing analysis of DNA obtained from formalin-fixed paraffin-embedded (FFPE) tumor tissues. The KRAS G12D mutation was then analyzed by ddPCR in FFPE tumors and stool-derived DNA from patients with this point mutation. Wild-type (WT) tumors, as determined by pyrosequencing, were included as controls; analysis of FFPE tissue and stool-derived DNA by ddPCR was performed for these patients as well.RESULTSAmong the total 70 patients included, KRAS mutations were detected by pyrosequencing in 32 (45.71%), whereas 38 (54.29%) had WT tumors. The frequency of KRAS mutations was higher in left-sided tumors (11 located in the right colon, 15 in the left, and 6 in the rectum). The predominant point mutation was KRAS G12D (14.29%, n = 10), which was more frequent in early-stage tumors (I-IIA, n = 7). In agreement with pyrosequencing results, the KRAS G12D mutation was detected by ddPCR in FFPE tumor-derived DNA, and only a residual number of mutated copies was found in WT controls. The KRAS G12D mutation was also detected in stool-derived DNA in 80% of all fecal samples from CRC patients with this point mutation.CONCLUSIONddPCR is a reliable and sensitive method to analyze KRAS G12D mutation in stool-derived DNA from CRC patients, especially at early stages. This non-invasive approach is potentially applicable to other relevant biomarkers for CRC management.展开更多
The system model of embedded reconfigurable system is discussed firstly, followed by the construction of the soft- ware and hardware platform. Based on the platform, the function trait and control models of digital si...The system model of embedded reconfigurable system is discussed firstly, followed by the construction of the soft- ware and hardware platform. Based on the platform, the function trait and control models of digital signal processing tasks are described. According the functional unit model and control model, a software/hardware integrating model with dataflow control infrastructure is constructed. The software/hardware partition is carded out in the experiments of ultrasonic signal processing, and the results show that the system model is practicable.展开更多
文摘对文本进行分词和词嵌入通常是中文命名实体识别的第一步,但中文的词与词之间没有明确的分界符,专业词及生僻词等未收录词(Out of Vocabulary,OOV)严重干扰了词向量的计算,基于词向量嵌入的模型性能极易受到分词效果的影响。同时现有模型大多使用循环神经网络,计算速度较慢,很难达到工业应用的要求。针对上述问题,构建了一个基于注意力机制和卷积神经网络的命名实体识别模型,即LAC-DGLU。针对分词依赖的问题,提出了一种基于局部注意力卷积(Local Attention Convolution,LAC)的字嵌入算法,减轻了模型对分词效果的依赖。针对计算速度较慢的问题,使用了一种带门结构的卷积神经网络,即膨胀门控线性单元(Dilated Gated Linear Unit,DGLU),提高了模型的计算速度。在多个数据集上的实验结果显示,该模型相比现有最优模型F1值提高了0.2%~2%,训练速度可以达到现有最优模型的1.4~1.9倍。
基金The National Public Benefit Research Foundation of China (No. 201111013-02)
文摘In order to resolve the conflict between the limited resources of embedded devices and the growing amount of massive image data to be shown, a solution for fast images rendering in embedded devices is proposed and implemented. First, an improved algorithm of a multi-resolution file-pyramid construction which is used for the organization of massive image data is presented. Then, a strategy, adopting technologies such as view-dependent levels of detail, target-tiles quick search and tiles seamless connection, is presented for fast scheduling and viewing of images. The results show that compared with the solution of multi-scale image representations based on wavelet, the proposed solution can improve the rendering speed, and the rendering speed does not depend on the image size, though it increases some data storage space. And the proposed solution is suitable for embedded devices and friendly experience.
文摘Optical fiber acceleration seismometer as an important instrument can offer high sensitivity, anti-jamming and non-touched advantage which has an extensive application field. Its signal processing ability will decide whole system’s performance to some extent because it will affect directly the factors such as resolving power, precision and dynamic range. The signal processing is usually realized by analog circuits which was more inferior in stability, flexibility and anti-jamming to digital processing system. A digital processing system of optical fiber acceleration seismometer has been designed based on the embedded system design scheme. Synthetic-heterodyne demodulation has been studied, and signal processing has been realized. The double processors of ARM and DSP are employed to implement respectively the system control and signal processing, and to provide the output interfaces such as LCD, DAC and Ethernet interface. This system can vary with the measured signal in real time and linearly, and its work frequency bandwidth is between 10Hz and 1kHz. The system has better anti-jamming ability and can work normally when the SNR is 40dB.
文摘Sign Writing, a writing system for sign language, is becoming a useful and convenient communication aid for people who are deaf. Principally, people who are deaf find it difficult to communicate with the hearing community and due to recent technological advancement they communicate amongst themselves and with the hearing community via text messaging on mobile phones. Existing messaging function is limited to writing based on the Roman alphabets or pictographic languages like Mandarin; writing in signs is deemed to be deficient in a mobile context. Hence, the aim of this paper is to examine the feasibility of writing and reading text messages in signs as an alternative communication mean besides Short Messaging Service （SMS）. Initial experimental results have significantly exemplified that sign writing gains well acceptance and is preferred among the hearing-impaired community to communicate within or between the hearing communities.
基金Supported by“Fondo de Investigaciones Sanitarias(FIS)-FEDER”,Ministry of Health,Spain,No.PI13/01924 to García-Olmo DRETIC Program of ISCIII-FEDER,No.RD12/0019/0035 to Olmedillas-López S
文摘AIMTo assess KRAS G12D mutation detection by droplet digital PCR (ddPCR) in stool-derived DNA from colorectal cancer (CRC) patients.METHODSIn this study, tumor tissue and stool samples were collected from 70 patients with stage I-IV CRC diagnosed by preoperative biopsy. KRAS mutational status was determined by pyrosequencing analysis of DNA obtained from formalin-fixed paraffin-embedded (FFPE) tumor tissues. The KRAS G12D mutation was then analyzed by ddPCR in FFPE tumors and stool-derived DNA from patients with this point mutation. Wild-type (WT) tumors, as determined by pyrosequencing, were included as controls; analysis of FFPE tissue and stool-derived DNA by ddPCR was performed for these patients as well.RESULTSAmong the total 70 patients included, KRAS mutations were detected by pyrosequencing in 32 (45.71%), whereas 38 (54.29%) had WT tumors. The frequency of KRAS mutations was higher in left-sided tumors (11 located in the right colon, 15 in the left, and 6 in the rectum). The predominant point mutation was KRAS G12D (14.29%, n = 10), which was more frequent in early-stage tumors (I-IIA, n = 7). In agreement with pyrosequencing results, the KRAS G12D mutation was detected by ddPCR in FFPE tumor-derived DNA, and only a residual number of mutated copies was found in WT controls. The KRAS G12D mutation was also detected in stool-derived DNA in 80% of all fecal samples from CRC patients with this point mutation.CONCLUSIONddPCR is a reliable and sensitive method to analyze KRAS G12D mutation in stool-derived DNA from CRC patients, especially at early stages. This non-invasive approach is potentially applicable to other relevant biomarkers for CRC management.
基金supported by the Research Project of “SUST Spring Bud”under contract number 2008BW2046 fron Shandong University of Science and Technology,China
文摘The system model of embedded reconfigurable system is discussed firstly, followed by the construction of the soft- ware and hardware platform. Based on the platform, the function trait and control models of digital signal processing tasks are described. According the functional unit model and control model, a software/hardware integrating model with dataflow control infrastructure is constructed. The software/hardware partition is carded out in the experiments of ultrasonic signal processing, and the results show that the system model is practicable.
