Chinese Term Extraction Based on PAT Tree

A new method of automatic Chinese term extraction is proposed based on Patricia （PAT） tree. Mutual information is calculated based on prefix searching in PAT tree of domain corpus to estimate the internal associative strength between Chinese characters in a string. It can improve the speed of term candidate extraction largely compared with methods based on domain corpus directly. Common collocation suffix, prefix bank are constructed and term part of speech （POS） composing rules are summarized to improve the precision of term extraction. Experiment results show that the F-measure is 74.97%.

Cross-Language Information Extraction and Auto Evaluation for OOV Term Translations

OOV term translation plays an important role in natural language processing. Although many researchers in the past have endeavored to solve the OOV term translation problems, but none existing methods offer definition or context information of OOV terms. Furthermore, non-existing methods focus on cross-language definition retrieval for OOV terms. Never the less, it has always been so difficult to evaluate the correctness of an OOV term translation without domain specific knowledge and correct references. Our English definition ranking method differentiate the types of OOV terms, and applies different methods for translation extraction. Our English definition ranking method also extracts multilingual context information and monolingual definitions of OOV terms. In addition, we propose a novel cross-language definition retrieval system for OOV terms. Never the less, we propose an auto re-evaluation method to evaluate the correctness of OOV translations and definitions. Our methods achieve high performances against existing methods.

Medical Knowledge Extraction and Analysis from Electronic Medical Records Using Deep Learning

Objectives Medical knowledge extraction (MKE) plays a key role in natural language processing (NLP) research in electronic medical records (EMR),which are the important digital carriers for recording medical activities of patients.Named entity recognition (NER) and medical relation extraction (MRE) are two basic tasks of MKE.This study aims to improve the recognition accuracy of these two tasks by exploring deep learning methods.Methods This study discussed and built two application scenes of bidirectional long short-term memory combined conditional random field (BiLSTM-CRF) model for NER and MRE tasks.In the data preprocessing of both tasks,a GloVe word embedding model was used to vectorize words.In the NER task,a sequence labeling strategy was used to classify each word tag by the joint probability distribution through the CRF layer.In the MRE task,the medical entity relation category was predicted by transforming the classification problem of a single entity into a sequence classification problem and linking the feature combinations between entities also through the CRF layer.Results Through the validation on the I2B2 2010 public dataset,the BiLSTM-CRF models built in this study got much better results than the baseline methods in the two tasks,where the F1-measure was up to 0.88 in NER task and 0.78 in MRE task.Moreover,the model converged faster and avoided problems such as overfitting.Conclusion This study proved the good performance of deep learning on medical knowledge extraction.It also verified the feasibility of the BiLSTM-CRF model in different application scenarios,laying the foundation for the subsequent work in the EMR field.

An Effective Concept Extraction Method for Improving Text Classification Performance

This paper presents anew way to extract concept that can beused to improve text classification per-formance (precision and recall). Thecomputational measure will be dividedinto two layers. The bottom layercalled document layer is concernedwith extracting the concepts of parti-cular document and the upper layercalled category layer is with findingthe description and subject concepts ofparticular category. The relevant im-plementation algorithm that dramatic-ally decreases the search space is dis-cussed in detail. The experiment basedon real-world data collected from Info-Bank shows that the approach is supe-rior to the traditional ones.

Meaningful String Extraction Based on Clustering for Improving Webpage Classification

Since webpage classification is different from traditional text classification with its irregular words and phrases,massive and unlabeled features,which makes it harder for us to obtain effective feature.To cope with this problem,we propose two scenarios to extract meaningful strings based on document clustering and term clustering with multi-strategies to optimize a Vector Space Model(VSM) in order to improve webpage classification.The results show that document clustering work better than term clustering in coping with document content.However,a better overall performance is obtained by spectral clustering with document clustering.Moreover,owing to image existing in a same webpage with document content,the proposed method is also applied to extract image meaningful terms,and experiment results also show its effectiveness in improving webpage classification.

Extractive Summarization Using Structural Syntax, Term Expansion and Refinement

This paper investigates a procedure developed and reports on experiments performed to studying the utility of applying a combined structural property of a text’s sentences and term expansion using WordNet [1] and a local thesaurus [2] in the selection of the most appropriate extractive text summarization for a particular document. Sentences were tagged and normalized then subjected to the Longest Common Subsequence (LCS) algorithm [3] [4] for the selection of the most similar subset of sentences. Calculated similarity was based on LCS of pairs of sentences that make up the document. A normalized score was calculated and used to rank sentences. A selected top subset of the most similar sentences was then tokenized to produce a set of important keywords or terms. The produced terms were further expanded into two subsets using 1) WorldNet;and 2) a local electronic dictionary/thesaurus. The three sets obtained (the original and the expanded two) were then re-cycled to further refine and expand the list of selected sentences from the original document. The process was repeated a number of times in order to find the best representative set of sentences. A final set of the top (best) sentences was selected as candidate sentences for summarization. In order to verify the utility of the procedure, a number of experiments were conducted using an email corpus. The results were compared to those produced by human annotators as well as to results produced using some basic sentences similarity calculation method. Produced results were very encouraging and compared well to those of human annotators and Jacquard sentences similarity.

Analysis and Simulation of Light Extraction of Light-Emitting Diodes:Simulation Efficiency

Two foundational factors （escape cone and transmissivity） about light extraction of light emitting diodes （LEDs） are discussed. According to these factors, a new process to simulate the light extraction of LEDs based on the Monte Carlo method has been provided. The improved method is to deal with the reflection and refraction of light （beam of light） at the interface between two mediums approximately. In addition, light extraction of traditional LEDs is simulated by different processes with the same structure and parameters. The results show that the reflection and refraction of light processed approximately are accurate enough for analyzing LEDs structure. This method saves much time and improves efficiency in the simulation of light extraction of LEDs.

Heavy metal chemistry in soils received long-term application of organic wastes

This study was carried out to understand the long-term effect of organic waste treatment on the fate of heavy metals originated from the organic wastes, together with examination of changes in soil properties. For this, the soils received three different organic wastes (municipal sewage sludge, alcohol fermentation processing sludge, pig manure compost) in three different rates (12.5, 25, 50 ton/ha/yr) for 7 years (1994 - 2000) were used. To see the long-term effect, plant growth study and soil examination were conducted twice in 2000 and 2010, respectively. There was no additional treatment of organic wastes for post ten years after ceasing organic waste treatment for seven years. Soil examination conducted in 2010 showed decreases in soil pH, EC, total nitrogen, organic matter, available phosphorus, exchangeable cations and heavy metal contents in all soils received organic wastes compared to the results obtained in 2000. Speciation of heavy metals in soil through sequential extraction showed that organically bound Cu was the dominant species in all treatment and exchangeable Cu was increased in the plots treated with municipal sewage sludge and alcohol fermentation processing sludge. organically bound Ni increased from 25% - 30% to 32% - 45% in 2010 inall treatment while Pb showed increase in carbonate form in all treatments. Zn existed mainly as sulfide and residual forms, showing increases in organically bound form in all treatment during post ten years.

基于改进麻雀搜索算法优化LSTM的滚动轴承故障诊断

为了对滚动轴承的工作状态及故障类别进行准确的诊断,本文采用长短时记忆(LSTM)神经网络作为分类器对滚动轴承数据集进行分类诊断。首先,从滚动轴承原始运行振动信号中提取时域和频域特征参数,组成具有高维特征参数的数据集;使用核主成分分析(KPCA)方法对高维特征集进行降维处理,选取重要性程度高的特征构成输入特征向量。然后,针对LSTM神经网络在滚动轴承故障诊断中存在的超参数难以确定的问题,提出一种基于自适应t分布策略的麻雀搜索算法优化LSTM神经网络的故障诊断方法(tSSA–LSTM)。最后,使用凯斯西储大学滚动轴承数据中心的数据进行故障诊断精度测试、泛化性能测试及噪声环境下故障诊断性能测试等多个仿真实验,并将本文提出的诊断模型与麻雀搜索算法优化长短时记忆神经网络(SSA–LSTM)、遗传算法优化长短时记忆神经网络(GA–LSTM)、粒子群算法优化长短时记忆神经网络(PSO–LSTM)及传统LSTM诊断模型进行对比。结果表明:tSSA可以更有效地对LSTM的隐含层神经元数量、周期次数、学习率等超参数进行合理优化;所提方法的平均诊断准确率达到98.86%,交叉验证平均诊断结果为98.57%;所提方法在噪声干扰下的故障诊断准确率也优于对比方法。因此,本文提出的tSSA–LSTM模型不仅可以更精准地诊断滚动轴承故障状态,而且具有更强的泛化能力及抗干扰能力,有效地提高了滚动轴承故障诊断的性能。

融合多源异构气象数据的光伏功率预测模型

高精度光伏功率预测对提高电力系统运行效率具有重要意义。光伏功率受多种因素影响,其中云层的变化是最主要的不确定因素。传统光伏功率预测方法没有充分考虑云的3维结构和气象要素对光伏功率的影响。因此,该文提出一种融合多源异构气象数据的多源变量光伏功率预测模型(MPPM)。MPPM的核心包括时空条件扩散模型(STCDM)、注意力堆叠LSTM网络(ASLSTM)和多维特征融合模块(MFFM)。STCDM模型通过对2维卫星云图进行精确预测,消除了云层边界处的模糊现象。ASLSTM模型则提取了3维天气研究与预报模式(WRF)气象要素特征。MFFM模块将2维卫星云图特征和3维WRF气象要素特征进行融合,以得到未来1 h光伏功率预测结果。该文分别利用STCDM模型和MPPM模型开展卫星云图预测实验和光伏功率预测实验。实验结果显示,STCDM模型预测1 h内卫星云图的结构相似性指数(SSIM)达到0.914,MPPM模型预测1 h内光伏功率的相关系数(CORR)达到0.949,优于所有对比算法。

EEMD与LSTM在轴承剩余寿命预测中的应用

剩余使用寿命(RUL)预测是实现装备健康管理与预测性维护的最主要技术手段之一,为了准确预测轴承的剩余使用寿命,提出了一种基于集合经验模态分解(EEMD)和长短时记忆网络(LSTM)的轴承剩余寿命预测方法。首先,对采集到的振动信号做时域、频域及时频分析,同时记录相应特征;进而,筛选特征,通过EEMD对振动信号予以分解并重构;最后,通过LSTM结合经过处理的信号构建健康特征指标。通过实验证明了该方法能有效的预测出轴承的剩余寿命,且有较高的预测精度。

超临界二氧化碳短时作用煤体裂隙演化CT/XRD实验研究

ECBM可以在提高煤层气开采效率的同时实现CO_(2)封存,推动“双碳”目标实现,在深部煤层气的开发中具有广泛的应用前景。煤层裂隙结构和分布规律决定了超临界二氧化碳(SC-CO_(2))的压裂效果,是提高煤层气开采效率的关键。长时作用下SC-CO_(2)与煤体接触产生的吸附膨胀和溶解萃取作用能够改变煤体裂隙形态,但压裂过程中SC-CO_(2)与煤体接触时间较短,短时作用下SC-CO_(2)的吸附膨胀及溶解萃取作用对煤体裂隙的影响尚未明确。因此开展了SC-CO_(2)短时作用下煤体的裂隙演化研究,通过CT扫描研究不同变质煤体裂隙随累计浸泡时间的变化规律;建立CT二维扫描图像的灰度分布函数并构建浸泡时间与裂隙演化规律的关系,定量表征煤体裂隙的变化。结合XRD实验研究浸泡时间对煤体物质成分变化的影响,明确SC-CO_(2)短时作用煤体裂隙演化规律的主要原因。结果表明:SC-CO_(2)短时作用下吸附膨胀作用会导致煤体裂隙收缩,溶解萃取作用使裂隙发生扩展;浸泡煤体过程中吸附膨胀和溶解萃取作用同时发生且强度随时间发生变化,在不同时间段内分别交替占据对煤体裂隙的主导作用;不同变质煤会影响吸附膨胀和溶解萃取作用的强度和主导时间。褐煤浸泡30 min时吸附膨胀占据主导作用,浸泡90~240 min溶解萃取作用增强并占据主导。烟煤浸泡30~90 min时吸附膨胀作用较强,浸泡90~240 min时溶解萃取占据主导。无烟煤浸泡30~150 min时吸附膨胀作用占据主导,浸泡150~240 min时溶解萃取占据主导。

Improved Term Weighting Technique for Automatic Web Page Classification

Automatic web page classification has become inevitable for web directories due to the multitude of web pages in the World Wide Web. In this paper an improved Term Weighting technique is proposed for automatic and effective classification of web pages. The web documents are represented as set of features. The proposed method selects and extracts the most prominent features reducing the high dimensionality problem of classifier. The proper selection of features among the large set improves the performance of the classifier. The proposed algorithm is implemented and tested on a benchmarked dataset. The results show the better performance than most of the existing term weighting techniques.

面向语法加权图文本的方面情感三元组抽取

方面情感三元组抽取包括方面抽取、意见抽取和方面情感分类3项任务,以管道方式解决该任务的研究方法无法利用元素之间的交互信息,同时也会造成错误传播和冗余训练。基于此,提出一种基于门控注意力和加权图文本的方面情感三元组抽取方法。采用双向长短时记忆网络学习句子的序列特征表示;利用门控注意力单元学习单词之间的线性联系;利用语法距离加权图卷积网络增强三元组元素之间的交互;利用网格标记推理策略预测三元组。在4个公开数据集上进行实验,结果表明:所提方法可以有效增强三元组元素之间的交互,提高三元组抽取的准确率;同时,所提方法的F1值分别为57.94%、70.54%、61.95%和67.66%,与基准模型相比均有所提高。

联合句法与位置信息的方面情感三元组抽取

为提高方面级情感三元组抽取任务的准确率,提出一种联合依存句法关系和位置偏移信息的抽取模型。在模型上下文编码中添加句法关系,结合图卷积网络捕获结构和结点属性信息,增强三元组要素之间的交互能力;在多任务学习部分加入相对位置偏移信息,充分挖掘方面-观点词对的关系,提高三元组要素抽取的精度。在4个基准英文数据集上的实验结果表明,该方法效果显著且优于其它基线模型。

融合BERT和双向长短时记忆网络的中文反讽识别研究

用户对微博热点话题进行评论时会使用反语、讽刺的修辞手法,其本身带有一定的情感倾向会对情感分析结果造成一定影响。因此该文主要针对中文微博评论进行反讽识别,构建了一个包含反语、讽刺和非反讽的三分类数据集,提出一个基于BERT和双向长短时记忆网络(BiLSTM)的模型BERT_BiLSTM。该模型通过BERT生成含有上下文信息的动态字向量,输入BiLSTM提取文本的深层反讽特征,在全连接层传入softmax对文本进行反讽识别。实验结果表示,在二分类和三分类数据集上,提出的BERT_BiLSTM模型与现有主流模型相比准确率和F1值均有明显提高。

生成式情报学术语自动抽取与多维关联知识挖掘研究

情报学术语承载了情报学科基础知识与核心概念。从概念维度梳理与分析情报学术语对推动学科发展、助力下游知识挖掘任务具有重要意义。面对数量快速增长的科技文献,自动术语抽取替代了人工筛选,但现有方法严重依赖大规模标注数据集,难以迁移至低资源场景。本文设计了一种生成式情报学术语抽取方法(generative term extraction for information science,GTX-IS),将传统基于序列标注的抽取式任务转化为序列到序列的生成式任务。结合小样本学习策略与有监督微调,提升面向特定任务的文本生成能力,能够在低资源有标签数据集场景下较为精准地抽取情报学术语。对于抽取结果,本文进一步开展了情报学领域术语发现及多维知识挖掘。综合运用全文科学计量与信息计量方法,从术语自身、术语间关联、时间信息等维度,对术语的出现频次、生命周期、共现信息等进行统计分析与知识挖掘。采用社会网络分析方法,结合时间维度特征,从术语角度出发,完善期刊的动态简介,探究情报学研究热点、演变历程和未来发展趋势。本文方法在术语抽取实验中的表现超越了全部13种主流生成式和抽取式模型,展现出较强的小样本学习能力,为领域信息抽取提供了新的思路。

基于状态划分和集成学习的轴承剩余使用寿命预测模型

针对滚动轴承剩余使用寿命(RUL)预测退化起始时间(DST)难以确定,以及单一寿命预测模型精度比较低的问题,提出了一种基于状态划分和集成学习模型的滚动轴承RUL预测方法。首先,提取了轴承振动信号的特征,利用滑动窗口不断更新3σ准则预警范围,结合连续触发机制自适应确定DST;然后,采用具有自适应噪声的完全集成经验模态分解(CEEMDAN)对退化阶段信号序列进行了自适应分解;最后,构建了集成学习模型,考虑分量的不同特性进行了多步滚动预测,融合预测结果得到了轴承RUL,采用滚动轴承XJTU-SY公开数据集进行了试验验证。研究结果表明:与基于长短时记忆神经网络(LSTM)、反向传播神经网络(BPNN)的预测方法相比,该方法预测结果的平均绝对误差分别降低了11.7%以及5.6%,相对均方根误差分别降低了12.2%以及10.7%,验证了该方法在轴承RUL预测中的有效性和优越性。

改进2DCNN时空特征提取的动作识别研究

基于深度学习的视频动作识别方法处理时间信息的方式主要有两种,一是利用光流表示相邻帧之间的运动信息,但其无法有效建模长程时间特征,二是利用3D卷积对时空信号进行混合建模,但其引入了大量的参数,导致内存消耗和计算量剧增.针对上述问题,本文提出了一种改进2D CNN时空特征提取的动作识别方法,在2D CNN中嵌入时空门控和动作注意力聚合(Spatial-temporal Gate and Motion Attention-aggregation,SGMA)模块增强其时空特征提取能力.SGMA包含时空动态门控和动作注意力聚合两个子模块,时空动态门控能够可视化各通道特征的运动比例因子并依此逐通道分离运动强相关特征和运动弱相关特征,动作注意力聚合利用运动强相关特征构建金字塔结构来提取不同时间跨度的运动特征,并使用注意力机制自适应聚合各时间跨度特征实现长程时间建模,运动弱相关特征经过2D卷积提取空间特征后融合动作注意力聚合模块的输出最终获得强有力的时空特征表达.在相同帧采样策略下,本文方法在Something-SomethingV1&V2验证集上的Top1准确度比基准TSM分别提高了4.4%和6.2%.