The increasing significance of text data in power system intelligence has highlighted the out-of-distribution(OOD)problem as a critical challenge,hindering the deployment of artificial intelligence(AI)models.In a clos...The increasing significance of text data in power system intelligence has highlighted the out-of-distribution(OOD)problem as a critical challenge,hindering the deployment of artificial intelligence(AI)models.In a closed-world setting,most AI models cannot detect and reject unexpected data,which exacerbates the harmful impact of the OOD problem.The high similarity between OOD and indistribution(IND)samples in the power system presents challenges for existing OOD detection methods in achieving effective results.This study aims to elucidate and address the OOD problem in power systems through a text classification task.First,the underlying causes of OOD sample generation are analyzed,highlighting the inherent nature of the OOD problem in the power system.Second,a novel method integrating the enhanced Mahalanobis distance with calibration strategies is introduced to improve OOD detection for text data in power system applications.Finally,the case study utilizing the actual text data from power system field operation(PSFO)is conducted,demonstrating the effectiveness of the proposed OOD detection method.Experimental results indicate that the proposed method outperformed existing methods in text OOD detection tasks within the power system,achieving a remarkable 21.03%enhancement of metric in the false positive rate at 95%true positive recall(FPR95)and a 12.97%enhancement in classi-fication accuracy for the mixed IND-OOD scenarios.展开更多
The performance of deep learning models is heavily reliant on the quality and quantity of train-ing data.Insufficient training data will lead to overfitting.However,in the task of alert-situation text classification,i...The performance of deep learning models is heavily reliant on the quality and quantity of train-ing data.Insufficient training data will lead to overfitting.However,in the task of alert-situation text classification,it is usually difficult to obtain a large amount of training data.This paper proposes a text data augmentation method based on masked language model(MLM),aiming to enhance the generalization capability of deep learning models by expanding the training data.The method em-ploys a Mask strategy to randomly conceal words in the text,effectively leveraging contextual infor-mation to predict and replace masked words based on MLM,thereby generating new training data.Three Mask strategies of character level,word level and N-gram are designed,and the performance of each Mask strategy under different Mask ratios is analyzed and studied.The experimental results show that the performance of the word-level Mask strategy is better than the traditional data augmen-tation method.展开更多
Data compression plays a key role in optimizing the use of memory storage space and also reducing latency in data transmission. In this paper, we are interested in lossless compression techniques because their perform...Data compression plays a key role in optimizing the use of memory storage space and also reducing latency in data transmission. In this paper, we are interested in lossless compression techniques because their performance is exploited with lossy compression techniques for images and videos generally using a mixed approach. To achieve our intended objective, which is to study the performance of lossless compression methods, we first carried out a literature review, a summary of which enabled us to select the most relevant, namely the following: arithmetic coding, LZW, Tunstall’s algorithm, RLE, BWT, Huffman coding and Shannon-Fano. Secondly, we designed a purposive text dataset with a repeating pattern in order to test the behavior and effectiveness of the selected compression techniques. Thirdly, we designed the compression algorithms and developed the programs (scripts) in Matlab in order to test their performance. Finally, following the tests conducted on relevant data that we constructed according to a deliberate model, the results show that these methods presented in order of performance are very satisfactory:- LZW- Arithmetic coding- Tunstall algorithm- BWT + RLELikewise, it appears that on the one hand, the performance of certain techniques relative to others is strongly linked to the sequencing and/or recurrence of symbols that make up the message, and on the other hand, to the cumulative time of encoding and decoding.展开更多
Clustering text data streams is an important issue in data mining community and has a number of applications such as news group filtering, text crawling, document organization and topic detection and tracing etc. Howe...Clustering text data streams is an important issue in data mining community and has a number of applications such as news group filtering, text crawling, document organization and topic detection and tracing etc. However, most methods are similarity-based approaches and only use the TF,IDF scheme to represent the semantics of text data and often lead to poor clustering quality. Recently, researchers argue that semantic smoothing model is more efficient than the existing TF,IDF scheme for improving text clustering quality. However, the existing semantic smoothing model is not suitable for dynamic text data context. In this paper, we extend the semantic smoothing model into text data streams context firstly. Based on the extended model, we then present two online clustering algorithms OCTS and OCTSM for the clustering of massive text data streams. In both algorithms, we also present a new cluster statistics structure named cluster profile which can capture the semantics of text data streams dynamically and at the same time speed up the clustering process. Some efficient implementations for our algorithms are also given. Finally, we present a series of experimental results illustrating the effectiveness of our technique.展开更多
Purpose： In the open science era, it is typical to share project-generated scientific data by depositing it in an open and accessible database. Moreover, scientific publications are preserved in a digital library arc...Purpose： In the open science era, it is typical to share project-generated scientific data by depositing it in an open and accessible database. Moreover, scientific publications are preserved in a digital library archive. It is challenging to identify the data usage that is mentioned in literature and associate it with its source. Here, we investigated the data usage of a government-funded cancer genomics project, The Cancer Genome Atlas（TCGA）, via a full-text literature analysis.Design/methodology/approach： We focused on identifying articles using the TCGA dataset and constructing linkages between the articles and the specific TCGA dataset. First, we collected 5,372 TCGA-related articles from Pub Med Central（PMC）. Second, we constructed a benchmark set with 25 full-text articles that truly used the TCGA data in their studies, and we summarized the key features of the benchmark set. Third, the key features were applied to the remaining PMC full-text articles that were collected from PMC.Findings： The amount of publications that use TCGA data has increased significantly since 2011, although the TCGA project was launched in 2005. Additionally, we found that the critical areas of focus in the studies that use the TCGA data were glioblastoma multiforme, lung cancer, and breast cancer; meanwhile, data from the RNA-sequencing（RNA-seq） platform is the most preferable for use.Research limitations： The current workflow to identify articles that truly used TCGA data is labor-intensive. An automatic method is expected to improve the performance.Practical implications： This study will help cancer genomics researchers determine the latest advancements in cancer molecular therapy, and it will promote data sharing and data-intensive scientific discovery.Originality/value： Few studies have been conducted to investigate data usage by governmentfunded projects/programs since their launch. In this preliminary study, we extracted articles that use TCGA data from PMC, and we created a link between the full-text articles and the source data.展开更多
Understanding the underlying goal behind a user＇s Web query has been proved to be helpful to improve the quality of search. This paper focuses on the problem of automatic identification of query types according to th...Understanding the underlying goal behind a user＇s Web query has been proved to be helpful to improve the quality of search. This paper focuses on the problem of automatic identification of query types according to the goals. Four novel entropy-based features extracted from anchor data and click-through data are proposed, and a support vector machines （SVM） classifier is used to identify the user＇s goal based on these features. Experi- mental results show that the proposed entropy-based features are more effective than those reported in previous work. By combin- ing multiple features the goals for more than 97% of the queries studied can be correctly identified. Besides these, this paper reaches the following important conclusions： First, anchor-based features are more effective than click-through-based features; Second, the number of sites is more reliable than the number of links; Third, click-distribution- based features are more effective than session-based ones.展开更多
In this paper, we analyze the complexity and entropy of different methods of data compression algorithms: LZW, Huffman, Fixed-length code (FLC), and Huffman after using Fixed-length code (HFLC). We test those algorith...In this paper, we analyze the complexity and entropy of different methods of data compression algorithms: LZW, Huffman, Fixed-length code (FLC), and Huffman after using Fixed-length code (HFLC). We test those algorithms on different files of different sizes and then conclude that: LZW is the best one in all compression scales that we tested especially on the large files, then Huffman, HFLC, and FLC, respectively. Data compression still is an important topic for research these days, and has many applications and uses needed. Therefore, we suggest continuing searching in this field and trying to combine two techniques in order to reach a best one, or use another source mapping (Hamming) like embedding a linear array into a Hypercube with other good techniques like Huffman and trying to reach good results.展开更多
Feature representation is one of the key issues in data clustering. The existing feature representation of scientific data is not sufficient, which to some extent affects the result of scientific data clustering. Ther...Feature representation is one of the key issues in data clustering. The existing feature representation of scientific data is not sufficient, which to some extent affects the result of scientific data clustering. Therefore, the paper proposes a concept of composite text description(CTD) and a CTD-based feature representation method for biomedical scientific data. The method mainly uses different feature weight algorisms to represent candidate features based on two types of data sources respectively, combines and finally strengthens the two feature sets. Experiments show that comparing with traditional methods, the feature representation method is more effective than traditional methods and can significantly improve the performance of biomedcial data clustering.展开更多
A method of fast data processing has been developed to rapidly obtain evolution of the electron density profile for a multichannel polarimeter-interferometer system（POLARIS）on J-TEXT. Compared with the Abel inversio...A method of fast data processing has been developed to rapidly obtain evolution of the electron density profile for a multichannel polarimeter-interferometer system（POLARIS）on J-TEXT. Compared with the Abel inversion method, evolution of the density profile analyzed by this method can quickly offer important information. This method has the advantage of fast calculation speed with the order of ten milliseconds per normal shot and it is capable of processing up to 1 MHz sampled data, which is helpful for studying density sawtooth instability and the disruption between shots. In the duration of a flat-top plasma current of usual ohmic discharges on J-TEXT, shape factor u is ranged from 4 to 5. When the disruption of discharge happens, the density profile becomes peaked and the shape factor u typically decreases to 1.展开更多
煤矿安全风险辨识文本包含丰富的风险特征描述与专家经验知识,深入挖掘这些文本对实现风险等级预测具有重要价值。针对风险辨识文本存在小样本、短文本及语义复杂问题,提出了一种融合类别描述与增强嵌入的煤矿安全风险预测模型。该方法...煤矿安全风险辨识文本包含丰富的风险特征描述与专家经验知识,深入挖掘这些文本对实现风险等级预测具有重要价值。针对风险辨识文本存在小样本、短文本及语义复杂问题,提出了一种融合类别描述与增强嵌入的煤矿安全风险预测模型。该方法在句子级嵌入维度对文本进行数据增强,有效扩充训练样本;通过构建风险类别描述引入煤矿领域知识,并利用注意力机制对风险类别描述进行动态融合,为煤矿安全风险样本补充专业知识;使用双向长短期记忆(Bidirectional Long Short-Term Memory,Bi-LSTM)网络与Mamba算法对原始文本特征进行深度提取,获取煤矿文本复杂情境下的核心特征;最后使用动态门控机制融合各模块特征,输出预测结果。研究表明,该模型在小规模煤矿风险辨识数据集上准确率和F1均有不错的表现,可基于煤矿安全风险辨识文本为煤矿安全风险等级预测提供支持。展开更多
为了更有效的利用已有数据资源,不造成科研设施的重复投资,数据共享越来越受到重视.NASA对地观测系统(EOS)提供了大量的包括MODIS在内的免费数据资源,为此,EOS Data Dumper(EDD)通过程序模拟EOS数据门户的正常下载流程,采用了先进的Web...为了更有效的利用已有数据资源,不造成科研设施的重复投资,数据共享越来越受到重视.NASA对地观测系统(EOS)提供了大量的包括MODIS在内的免费数据资源,为此,EOS Data Dumper(EDD)通过程序模拟EOS数据门户的正常下载流程,采用了先进的Web页面文本信息捕捉技术,实现定时自动下载研究区的全部EOS免费数据,并通过免费的DIAL系统,向互联网重新发布,实现复杂的基于时空的数据查询.从技术角度详细介绍了EDD的项目背景与意义、实现方案。展开更多
The rapid development of the Internet brings a variety of original information including text information, audio information, etc. However, it is difficult to find the most useful knowledge rapidly and accurately beca...The rapid development of the Internet brings a variety of original information including text information, audio information, etc. However, it is difficult to find the most useful knowledge rapidly and accurately because of its huge number. Automatic text classification technology based on machine learning can classify a large number of natural language documents into the corresponding subject categories according to its correct semantics. It is helpful to grasp the text information directly. By learning from a set of hand-labeled documents,we obtain the traditional supervised classifier for text categorization(TC). However, labeling all data by human is labor intensive and time consuming. To solve this problem, some scholars proposed a semi-supervised learning method to train classifier, but it is unfeasible for various kinds and great number of Web data since it still needs a part of hand-labeled data. In 2012, Li et al. invented a fully automatic categorization approach for text(FACT)based on supervised learning, where no manual labeling efforts are required. But automatically labeling all data can bring noise into experiment and cause the fact that the result cannot meet the accuracy requirement. We put forward a new idea that part of data with high accuracy can be automatically tagged based on the semantic of category name, then a semi-supervised way is taken to train classifier with both labeled and unlabeled data,and ultimately a precise classification of massive text data can be achieved. The empirical experiments show that the method outperforms the supervised support vector machine(SVM) in terms of both F1 performance and classification accuracy in most cases. It proves the effectiveness of the semi-supervised algorithm in automatic TC.展开更多
基金supported in part by the Science and Technology Project of the State Grid East China Branch(No.520800230008).
文摘The increasing significance of text data in power system intelligence has highlighted the out-of-distribution(OOD)problem as a critical challenge,hindering the deployment of artificial intelligence(AI)models.In a closed-world setting,most AI models cannot detect and reject unexpected data,which exacerbates the harmful impact of the OOD problem.The high similarity between OOD and indistribution(IND)samples in the power system presents challenges for existing OOD detection methods in achieving effective results.This study aims to elucidate and address the OOD problem in power systems through a text classification task.First,the underlying causes of OOD sample generation are analyzed,highlighting the inherent nature of the OOD problem in the power system.Second,a novel method integrating the enhanced Mahalanobis distance with calibration strategies is introduced to improve OOD detection for text data in power system applications.Finally,the case study utilizing the actual text data from power system field operation(PSFO)is conducted,demonstrating the effectiveness of the proposed OOD detection method.Experimental results indicate that the proposed method outperformed existing methods in text OOD detection tasks within the power system,achieving a remarkable 21.03%enhancement of metric in the false positive rate at 95%true positive recall(FPR95)and a 12.97%enhancement in classi-fication accuracy for the mixed IND-OOD scenarios.
基金Supported by the Humanities and Social Sciences Research Project of the Ministry of Education(No.22YJA840004).
文摘The performance of deep learning models is heavily reliant on the quality and quantity of train-ing data.Insufficient training data will lead to overfitting.However,in the task of alert-situation text classification,it is usually difficult to obtain a large amount of training data.This paper proposes a text data augmentation method based on masked language model(MLM),aiming to enhance the generalization capability of deep learning models by expanding the training data.The method em-ploys a Mask strategy to randomly conceal words in the text,effectively leveraging contextual infor-mation to predict and replace masked words based on MLM,thereby generating new training data.Three Mask strategies of character level,word level and N-gram are designed,and the performance of each Mask strategy under different Mask ratios is analyzed and studied.The experimental results show that the performance of the word-level Mask strategy is better than the traditional data augmen-tation method.
文摘Data compression plays a key role in optimizing the use of memory storage space and also reducing latency in data transmission. In this paper, we are interested in lossless compression techniques because their performance is exploited with lossy compression techniques for images and videos generally using a mixed approach. To achieve our intended objective, which is to study the performance of lossless compression methods, we first carried out a literature review, a summary of which enabled us to select the most relevant, namely the following: arithmetic coding, LZW, Tunstall’s algorithm, RLE, BWT, Huffman coding and Shannon-Fano. Secondly, we designed a purposive text dataset with a repeating pattern in order to test the behavior and effectiveness of the selected compression techniques. Thirdly, we designed the compression algorithms and developed the programs (scripts) in Matlab in order to test their performance. Finally, following the tests conducted on relevant data that we constructed according to a deliberate model, the results show that these methods presented in order of performance are very satisfactory:- LZW- Arithmetic coding- Tunstall algorithm- BWT + RLELikewise, it appears that on the one hand, the performance of certain techniques relative to others is strongly linked to the sequencing and/or recurrence of symbols that make up the message, and on the other hand, to the cumulative time of encoding and decoding.
基金Supported by the National Natural Science Foundation of China under Grant Nos.60573097,60703111,60773198the Natural Science Foundation of Guangdong Province under Grant No.06104916+1 种基金the Specialized Research Foundation for the Doctoral Program of Higher Education under Grant No.20050558017the Program for New Century Excellent Talents in University of China under Grant No.NCET-06-0727.
文摘Clustering text data streams is an important issue in data mining community and has a number of applications such as news group filtering, text crawling, document organization and topic detection and tracing etc. However, most methods are similarity-based approaches and only use the TF,IDF scheme to represent the semantics of text data and often lead to poor clustering quality. Recently, researchers argue that semantic smoothing model is more efficient than the existing TF,IDF scheme for improving text clustering quality. However, the existing semantic smoothing model is not suitable for dynamic text data context. In this paper, we extend the semantic smoothing model into text data streams context firstly. Based on the extended model, we then present two online clustering algorithms OCTS and OCTSM for the clustering of massive text data streams. In both algorithms, we also present a new cluster statistics structure named cluster profile which can capture the semantics of text data streams dynamically and at the same time speed up the clustering process. Some efficient implementations for our algorithms are also given. Finally, we present a series of experimental results illustrating the effectiveness of our technique.
基金supported by the National Population and Health Scientific Data Sharing Program of Chinathe Knowledge Centre for Engineering Sciences and Technology (Medical Centre)the Fundamental Research Funds for the Central Universities (Grant No.: 13R0101)
文摘Purpose： In the open science era, it is typical to share project-generated scientific data by depositing it in an open and accessible database. Moreover, scientific publications are preserved in a digital library archive. It is challenging to identify the data usage that is mentioned in literature and associate it with its source. Here, we investigated the data usage of a government-funded cancer genomics project, The Cancer Genome Atlas（TCGA）, via a full-text literature analysis.Design/methodology/approach： We focused on identifying articles using the TCGA dataset and constructing linkages between the articles and the specific TCGA dataset. First, we collected 5,372 TCGA-related articles from Pub Med Central（PMC）. Second, we constructed a benchmark set with 25 full-text articles that truly used the TCGA data in their studies, and we summarized the key features of the benchmark set. Third, the key features were applied to the remaining PMC full-text articles that were collected from PMC.Findings： The amount of publications that use TCGA data has increased significantly since 2011, although the TCGA project was launched in 2005. Additionally, we found that the critical areas of focus in the studies that use the TCGA data were glioblastoma multiforme, lung cancer, and breast cancer; meanwhile, data from the RNA-sequencing（RNA-seq） platform is the most preferable for use.Research limitations： The current workflow to identify articles that truly used TCGA data is labor-intensive. An automatic method is expected to improve the performance.Practical implications： This study will help cancer genomics researchers determine the latest advancements in cancer molecular therapy, and it will promote data sharing and data-intensive scientific discovery.Originality/value： Few studies have been conducted to investigate data usage by governmentfunded projects/programs since their launch. In this preliminary study, we extracted articles that use TCGA data from PMC, and we created a link between the full-text articles and the source data.
基金the Tianjin Applied Fundamental Research Plan (07JCYBJC14500)
文摘Understanding the underlying goal behind a user＇s Web query has been proved to be helpful to improve the quality of search. This paper focuses on the problem of automatic identification of query types according to the goals. Four novel entropy-based features extracted from anchor data and click-through data are proposed, and a support vector machines （SVM） classifier is used to identify the user＇s goal based on these features. Experi- mental results show that the proposed entropy-based features are more effective than those reported in previous work. By combin- ing multiple features the goals for more than 97% of the queries studied can be correctly identified. Besides these, this paper reaches the following important conclusions： First, anchor-based features are more effective than click-through-based features; Second, the number of sites is more reliable than the number of links; Third, click-distribution- based features are more effective than session-based ones.
文摘In this paper, we analyze the complexity and entropy of different methods of data compression algorithms: LZW, Huffman, Fixed-length code (FLC), and Huffman after using Fixed-length code (HFLC). We test those algorithms on different files of different sizes and then conclude that: LZW is the best one in all compression scales that we tested especially on the large files, then Huffman, HFLC, and FLC, respectively. Data compression still is an important topic for research these days, and has many applications and uses needed. Therefore, we suggest continuing searching in this field and trying to combine two techniques in order to reach a best one, or use another source mapping (Hamming) like embedding a linear array into a Hypercube with other good techniques like Huffman and trying to reach good results.
基金supported by the Agridata,the sub-program of National Science and Technology Infrastructure Program(Grant No.2005DKA31800)
文摘Feature representation is one of the key issues in data clustering. The existing feature representation of scientific data is not sufficient, which to some extent affects the result of scientific data clustering. Therefore, the paper proposes a concept of composite text description(CTD) and a CTD-based feature representation method for biomedical scientific data. The method mainly uses different feature weight algorisms to represent candidate features based on two types of data sources respectively, combines and finally strengthens the two feature sets. Experiments show that comparing with traditional methods, the feature representation method is more effective than traditional methods and can significantly improve the performance of biomedcial data clustering.
基金supported by the National Magnetic Confinement Fusion Science Program of China(Nos.2014GB106000,2014GB106002,and2014GB106003)National Natural Science Foundation of China(Nos.11275234,11375237 and 11505238)Scientific Research Grant of Hefei Science Center of CAS(No.2015SRG-HSC010)
文摘A method of fast data processing has been developed to rapidly obtain evolution of the electron density profile for a multichannel polarimeter-interferometer system（POLARIS）on J-TEXT. Compared with the Abel inversion method, evolution of the density profile analyzed by this method can quickly offer important information. This method has the advantage of fast calculation speed with the order of ten milliseconds per normal shot and it is capable of processing up to 1 MHz sampled data, which is helpful for studying density sawtooth instability and the disruption between shots. In the duration of a flat-top plasma current of usual ohmic discharges on J-TEXT, shape factor u is ranged from 4 to 5. When the disruption of discharge happens, the density profile becomes peaked and the shape factor u typically decreases to 1.
文摘煤矿安全风险辨识文本包含丰富的风险特征描述与专家经验知识,深入挖掘这些文本对实现风险等级预测具有重要价值。针对风险辨识文本存在小样本、短文本及语义复杂问题,提出了一种融合类别描述与增强嵌入的煤矿安全风险预测模型。该方法在句子级嵌入维度对文本进行数据增强,有效扩充训练样本;通过构建风险类别描述引入煤矿领域知识,并利用注意力机制对风险类别描述进行动态融合,为煤矿安全风险样本补充专业知识;使用双向长短期记忆(Bidirectional Long Short-Term Memory,Bi-LSTM)网络与Mamba算法对原始文本特征进行深度提取,获取煤矿文本复杂情境下的核心特征;最后使用动态门控机制融合各模块特征,输出预测结果。研究表明,该模型在小规模煤矿风险辨识数据集上准确率和F1均有不错的表现,可基于煤矿安全风险辨识文本为煤矿安全风险等级预测提供支持。
文摘为了更有效的利用已有数据资源,不造成科研设施的重复投资,数据共享越来越受到重视.NASA对地观测系统(EOS)提供了大量的包括MODIS在内的免费数据资源,为此,EOS Data Dumper(EDD)通过程序模拟EOS数据门户的正常下载流程,采用了先进的Web页面文本信息捕捉技术,实现定时自动下载研究区的全部EOS免费数据,并通过免费的DIAL系统,向互联网重新发布,实现复杂的基于时空的数据查询.从技术角度详细介绍了EDD的项目背景与意义、实现方案。
基金the National Key Technology Research and Development Program of China(No.2015BAH13F01)the Beijing Natural Science Foundation(No.4152007)
文摘The rapid development of the Internet brings a variety of original information including text information, audio information, etc. However, it is difficult to find the most useful knowledge rapidly and accurately because of its huge number. Automatic text classification technology based on machine learning can classify a large number of natural language documents into the corresponding subject categories according to its correct semantics. It is helpful to grasp the text information directly. By learning from a set of hand-labeled documents,we obtain the traditional supervised classifier for text categorization(TC). However, labeling all data by human is labor intensive and time consuming. To solve this problem, some scholars proposed a semi-supervised learning method to train classifier, but it is unfeasible for various kinds and great number of Web data since it still needs a part of hand-labeled data. In 2012, Li et al. invented a fully automatic categorization approach for text(FACT)based on supervised learning, where no manual labeling efforts are required. But automatically labeling all data can bring noise into experiment and cause the fact that the result cannot meet the accuracy requirement. We put forward a new idea that part of data with high accuracy can be automatically tagged based on the semantic of category name, then a semi-supervised way is taken to train classifier with both labeled and unlabeled data,and ultimately a precise classification of massive text data can be achieved. The empirical experiments show that the method outperforms the supervised support vector machine(SVM) in terms of both F1 performance and classification accuracy in most cases. It proves the effectiveness of the semi-supervised algorithm in automatic TC.
