基于散列技术的快速子串归并算法

用统计方法研究东西方语言的多词单元问题和东方语言的未登录词问题时需要删除同频子串(子串归并)．传统的子串归并算法时间复杂度为O(n^2),在大规模语料库的处理中效率低下．提出一种基于散列技术的时间复杂度为O(n^2)的子串归并算法,并用数学方法证明其与O(n^2)复杂度的算法等价,即输入相同时输出也相同．不同规模语料上的实验结果表明新算法能够大大缩短子串归并所需时间,适用于大规模语料库的处理．     

S12交换机EC7.4版用户分组

从源的分析、字冠分析、任务单元定义3个方面,阐述了S12交换机EC7.4版的用户分组问题。     

Subshifts on Infinite Alphabets and Their Entropy

We analyze symbolic dynamics to infinite alphabets by endowing the alphabet with the cofinite topology. The topological entropy is shown to be equal to the supremum of the growth rate of the complexity function with respect to finite subalphabets. For the case of topological Markov chains induced by countably infinite graphs, our approach yields the same entropy as the approach of Gurevich We give formulae for the entropy of countable topological Markov chains in terms of the spectral radius in l2.     

构建单字词表识别未登录词的方法

针对目前中文分词技术主要依赖于常用词词典,而词典对未登录词识别率较低的问题,提出一种用双词典识别未登录词的方法,即构建一个常用词词典和一个单字词词典,二者相互结合进行分词,有效解决了对未登录词识别效率偏低的问题.实验表明,采用构建单字词表法对未登录词的识别准确率可达90%以上.     

Reduced Words and Plane Partitions

Let w ₀ be the element of maximal length in thesymmetric group S _n , and let Red(w ₀) bethe set of all reduced words for w ₀. We prove the identity which generalizes Stanley's [20] formula forthe cardinality of Red(w ₀), and Macdonald's [11] formula .Our approach uses anobservation, based on a result by Wachs [21], that evaluation of certainspecializations of Schubert polynomials is essentially equivalent toenumeration of plane partitions whose parts are bounded from above. Thus,enumerative results for reduced words can be obtained from the correspondingstatements about plane partitions, and vice versa. In particular, identity(*) follows from Proctor's [14] formula for the number of planepartitions of a staircase shape, with bounded largest part.Similar results are obtained for other permutations and shapes;q-analogues are also given.     

基于KNN的Android智能手机微信取证方法

针对微信数据多,无法从中快速找到与案件相关数据的问题,提出了一种基于KNN(k-nearest neighbor)算法的Android智能手机微信取证方法。引入词语相似度计算会话间的距离,将微信会话表示成特征词的向量,用KNN算法对会话进行分类,迅速找到与犯罪有关的聊天内容,并通过实验验证了该方法的可行性与准确性。     

Brugnanin,a new syn-2,3-dihydrobenzofuran neolignan dioate from the mangrove <Emphasis Type="Italic">Bruguiera gymnorrhiza</Emphasis>

Brugnanin (1), a neolignan dioate, was isolated from a mangrove plant Bruguiera gymnorrhiza. Based on spectroscopic interpretation of MS, IR, and NMR data, 1 was elucidated as (7R*,8S*,E)-3-hydroxy-5,5′-dimethoxy-7-O-4′,8-3′-neolignan-7′-ene-9,9′-dioic acid dioctadecyl ester. MTT assay showed that 1 had weak inhibitory activity against growth of CNE-1 nasopharyngeal carcinoma cell line. Published in Khimiya Prirodnykh Soedinenii, No. 2, pp. 147–149, March–April, 2008.     

基于网页特征的特征词提取技术

特征词提取是一项提炼整个web页面内容的实用技术,同时也为文本分类,信息抽取应用提供了技术支持.在web页面内容上,利用段落间语义关系划分出网页内容的篇章结构,并以此为基础使用网页的元数据和特殊标签,设计了一个特征词的加权函数,综合考虑了词频、词长和位置因子,最后,实验对比了各类位置因子对系统的贡献度.实验结果表明,改进方法的F1值比传统的TFIDF提取技术提高了15.5%,其中,位置因子中的标题,关键词和摘要因素对系统的贡献最大.     

基于主题模型的中文词义归纳

词义归纳是在给定包含多义词语料的条件下,识别出多义词词义的过程,通常是采用聚类的方法.本文提出了基于主题模型的方法来解决中文词义归纳问题,基于主题模型的词义归纳方法关键之处在于使用文档的主题概率分布来推断多义词的词义分布.实验结果表明,本文方法在测试数据上获得了77.58%FScore值.