深度学习与关联模式挖掘融合的查询扩展

本文提出一种深度学习与关联模式挖掘融合的查询扩展模型.该模型采用基于Copulas函数的支持度-置信度评价框架挖掘初检伪相关反馈文档集中扩展词,构建统计扩展词集,利用深度学习工具对初检文档集进行词向量语义学习训练得到词向量扩展词集,将统计扩展词集和词向量扩展词集融合得到最终扩展词.该模型不仅考虑来自统计分析与挖掘的扩展词与原查询间的关联信息,还考虑扩展词在文档中的上下文语义信息,扩展词质量得到较好地改善.在NTCIR-5 CLIR语料的实验结果表明,本文扩展模型能提高信息检索性能,其MAP和P@5平均增幅高于近年现有同类查询扩展方法.本文扩展模型可用于跨语言检索系统,以提高其性能.     

多查询相关的排序支持向量机融合算法

排序学习是目前信息检索与机器学习领域研究的热点问题.现有排序学习算法在学习时把训练样本集中的所有查询及其相关文档等同对待,忽视了查询之间的差异,影响了排序模型的性能.对查询之间的差异进行描述,并在训练过程中考虑这种差异,提出一种基于有监督学习的融合多个与查询相关排序子模型的方法.该方法为每一个查询及其相关文档建立一个子排序模型,并将子排序模型的输出进行向量化表示,将多个查询相关的排序模型转化为体现查询差异的特征数据,实现多排序模型的集成.以排序支持向量机为例,在查询级和样本级建立新的损失函数作为优化目标,并利用此损失函数调节不同查询产生损失之间的权重,提出多查询相关的排序支持向量机融合算法.在文档检索和网页检索中的实验结果表明,使用多查询相关的排序支持向量机融合算法可以取得比传统排序学习模型更好的性能.     

基于语词抽取与负关联规则挖掘的信息检索

将语词抽取、负关联规则挖掘和查询扩展技术应用于信息检索,提出一种基于语词抽取与负关联规则挖掘融合的信息检索系统模型及其算法.详细论述模型的设计思想、各模块的功能,以及模型的理论分析和检索算法.该模型能够将语词抽取、负关联规则挖掘和查询扩展三种技术融合,对初检文档集进行有效地处理,得到高质量的与原查询词相关的扩展词,和原查询组合成新查询,再进行二次检索,有效地解决了词不匹配的问题.实验结果表明,该模型有效,能改善和提高信息检索性能     

基于文档实例的中文信息检索

传统的信息检索系统基于关键词建立索引并进行信息检索.这些系统存在查询返回文档集大、准确率低和普通用户不便于构造查询等不足.为此,该文提出基于文档实例的信息检索,即以已有文档作为样本,在文档库中检索与样本文档相似的所有文档.文中给出了基于文档实例的中文信息检索的解决方法和实现技术.初步实验结果表明该方法是行之有效的.     

基于贝叶斯和cGAN的甲状腺结节生成方法

为克服小样本医学图像数据集对训练强大的超声图像中甲状腺结节的自动分割或分类的深度学习模型的限制,提出一种基于贝叶斯神经网络和条件生成对抗网络(cGAN)的数据生成方法.构建cGAN并以修改的结节掩膜作为条件,通过在真实样本上训练生成器生成具有不同特征的结节.使用贝叶斯神经网络识别出最有价值的生成样本,并用其构建新的数据集训练模型进行结节的分割和分类.实验结果表明,提出方法仅使用约35％的真实样本即可实现令人满意的分割和分类性能.     

一种改进的XML向量空间模型及其近似匹配算法

XML已成为网上数据交换和存储的标准,然而现有的XML文档模型存在很多问题.综合了数据挖掘和信息检索两方面的技术,对传统的向量空间模型加以扩展,提出一种新的基于语义和支持度的XML向量空间模型并给出其生成算法,该模型抽取文档集的频繁路径作为特征,然后利用XML本身的语义特性,对标签之间作ontology判断.最后,基于该模型提出一种XML近似查询算法,将文档矩阵从改进的VSM空间映射到潜在语义空间,并在转换后的空间获得近似查询结果并排序.对提出的方法进行理论分析和实验验证,得到满意的结果.     

基于多表达的第一阶段语义检索模型

当前,信息检索系统通常采用“检索+重排序”的多级流水线架构。基于稠密表示的检索模型已经被逐渐应用到第一阶段检索中,并展现出了相比传统的稀疏向量空间模型更好的性能。考虑到第一阶段检索所需的高效性,大多数情况下这些模型的基本架构都采用双编码器（bi-encoder）结构。对查询和文档进行独立的编码,分别得到一个稠密表示向量,然后基于获得的查询和文档表示使用简单的相似度函数计算查询-文档对的得分。然而,在编码文档的过程中查询是不可知的,而且文档相比查询而言通常包含更多的主题信息,因此这种简单的单表示模型可能会造成严重的文档信息丢失。为了解决这个问题,设计了一种新的语义检索方法 MDR(multi-representation dense retrieval),将文档编码成多个稠密向量表示。同时,该方法引入覆盖率（coverage）机制来保证多个向量之间的差异性,从而能够覆盖文档中不同主题的信息。为了评估模型性能,在MS MARCO数据集上进行了段落排序和文档排序任务,实验结果证明了MDR方法的有效性。     

结合句子级别检索的信息检索模型

查询词之间的距离较为接近的文档,相关的可能性更大,将这种距离信息用于信息检索模型的构造可有效提高检索的性能。然而直接估计查询词在文档中的距离需要大量的训练文本,且计算复杂度高。该文提出了一种结合句子级别检索的信息检索模型,将文档分为若干个窗口,通过计算句子和查询的相关度考察查询词在给定窗口中的共现性,该方法可增大那些查询词彼此靠近的文档的相关度,从而使得检索模型可返回更为相关的文档。标准数据集上的实验结果表明所提出的模型可以取得较好的性能。     

适应文档检索的半监督多样本排序学习算法

针对标记训练集不足的问题,提出了一种协同训练的多样本排序学习算法,从无标签数据挖掘隐含的排序信息。算法使用了两类多样本排序学习机,从当前已有的标记数据集分别构造两个不同的排序函数。相应地,每一个无标签查询都有两个不同的文档排列,由似然损失来计算这两个排列的相似性,为那些文档排列相似度低的查询贴上标签,使两个多样本排序学习机新增了训练数据。在排序学习公开数据集LETOR上的实验结果证实,协同训练的排序算法很有效。另外,还讨论了标注比例对算法的影响。     

直接优化性能指标的多排序模型融合方法

现有排序学习算法忽视了查询之间的差异,在建立排序模型的过程中等同对待训练样本集中的所有查询及其相关文档,影响了排序模型的性能.文中描述了查询之间的差异,并在训练过程中考虑查询之间的差异,提出了一种基于有监督学习的多排序模型融合方法.这种方法首先使用每一个查询及其相关文档训练出子排序模型,并将每一个子排序模型的输出转化为体现查询差异的特征数据,使用监督学习方法,实现了多排序模型的融合.更进一步,针对排序问题的特性,文中提出了一种直接优化排序性能的融合函数融合子排序模型,使用梯度上升方法优化其下界函数.文中证明了直接优化排序性能的融合函数融合子排序模型的性能优于子排序模型线性合并的性能.基于较大规模真实数据应用的实验结果表明,直接优化性能指标的多排序模型融合方法可以比传统排序学习模型具有更好的排序性能.     

Measuring ranked list robustness for query performance prediction

We introduce the notion of ranking robustness, which refers to a property of a ranked list of documents that indicates how stable the ranking is in the presence of uncertainty in the ranked documents. We propose a statistical measure called the robustness score to quantify this notion. Our initial motivation for measuring ranking robustness is to predict topic difficulty for content-based queries in the ad-hoc retrieval task. Our results demonstrate that the robustness score is positively and consistently correlation with average precision of content-based queries across a variety of TREC test collections. Though our focus is on prediction under the ad-hoc retrieval task, we observe an interesting negative correlation with query performance when our technique is applied to named-page finding queries, which are a fundamentally different kind of queries. A side effect of this different behavior of the robustness score between the two types of queries is that the robustness score is also found to be a good feature for query classification.      

基于迭代方法的多层Markov网络信息检索模型

查询扩展是提高检索效果的有效方法,传统的查询扩展方法大都以单个查询词的相关性来扩展查询词,没有充分考虑词项之间、文档之间以及查询之间的相关性,使得扩展效果不佳。针对此问题,该文首先通过分别构造词项子空间和文档子空间的Markov网络,用于提取出最大词团和最大文档团,然后根据词团与文档团的映射关系将词团分为文档依赖和非文档依赖词团,并构建基于文档团依赖的Markov网络检索模型做初次检索,从返回的检索结果集合中构造出查询子空间的Markov网络,用于提取出最大查询团,最后,采用迭代的方法计算文档与查询的相关概率,并构建出最终的基于迭代方法的多层Markov网络信息检索模型。实验结果表明 该文的模型能较好地提高检索效果。     

基于多层Markov网络的信息检索模型

随着信息检索技术的不断发展,挖掘更加有效的信息来提高检索精度成为研究热点,已有的研究表明在检索过程中有效地融合各种信息将得到更好的检索效果。对一个具体查询而言,可以充分利用与已有查询的相关性、词语相关性和文档相关性等信息进行查询扩展和重构。基于这种思路,该文分别构造查询网络、词网络和文档网络,提出了多层Markov网络的信息检索模型,模型可以融合词间关系、文档间关系和查询间关系,为了有效降低计算量,给出了基于团计算模型。在标准数据集上的实验表明该文的模型能够有效融合三类信息,并较大幅度地提高检索效果。     

Training query filtering for semi-supervised learning to rank with pseudo labels

Semi-supervised learning is a machine learning paradigm that can be applied to create pseudo labels from unlabeled data for learning a ranking model, when there is only limited or no training examples available. However, the effectiveness of semi-supervised learning in information retrieval (IR) can be hindered by the low quality pseudo labels, hence the need for the training query filtering that removes the low quality queries. In this paper, we assume two application scenarios with respect to the availability of human labels. First, for applications without any labeled data available, a clustering-based approach is proposed to select the high quality training queries. This approach selects the training queries following the empirical observation that the relevant documents of high quality training queries are highly coherent. Second, for applications with limited labeled data available, a classification-based approach is proposed. This approach learns a weak classifier to predict the retrieval performance gain of a given training query by making use of query features. The queries with high performance gains are selected for the following transduction process to create the pseudo labels for learning to rank algorithms. Experimental results on the standard LETOR dataset show that our proposed approaches outperform the strong baselines.     

Query Expansion for Effective Retrieval Results of Hindi–English Cross-Lingual IR

Information retrieval (IR) is the science of identifying documents or sub-documents from a collection of information or database. The collection of information does not necessarily be available in only one language as information does not depend on languages. Monolingual IR is the process of retrieving information in query language whereas cross-lingual information retrieval (CLIR) is the process of retrieving information in a language that differs from query language. In current scenario, there is a strong demand of CLIR system because it allows the user to expand the international scope of searching a relevant document. As compared to monolingual IR, one of the biggest problems of CLIR is poor retrieval performance that occurs due to query mismatching, multiple representations of query terms and untranslated query terms. Query expansion (QE) is the process or technique of adding related terms to the original query for query reformulation. Purpose of QE is to improve the performance and quality of retrieved information in CLIR system. In this paper, QE has been explored for a Hindi–English CLIR in which Hindi queries are used to search English documents. We used Okapi BM25 for documents ranking, and then by using term selection value, translated queries have been expanded. All experiments have been performed using FIRE 2012 dataset. Our result shows that the relevancy of Hindi–English CLIR can be improved by adding the lowest frequency term.     

Irrelevance reduction with locality-sensitive hash learning for efficient cross-media retrieval

Cross-media retrieval is an imperative approach to handle the explosive growth of multimodal data on the web. However, existing approaches to cross-media retrieval are computationally expensive due to high dimensionality. To efficiently retrieve in multimodal data, it is essential to reduce the proportion of irrelevant documents. In this paper, we propose a fast cross-media retrieval approach (FCMR) based on locality-sensitive hashing (LSH) and neural networks. One modality of multimodal information is projected by LSH algorithm to cluster similar objects into the same hash bucket and dissimilar objects into different ones and then another modality is mapped into these hash buckets using hash functions learned through neural networks. Once given a textual or visual query, it can be efficiently mapped to a hash bucket in which objects stored can be near neighbors of this query. Experimental results show that, in the set of the queries’ near neighbors obtained by the proposed method, the proportions of relevant documents can be much boosted, and it indicates that the retrieval based on near neighbors can be effectively conducted. Further evaluations on two public datasets demonstrate the efficacy of the proposed retrieval method compared to the baselines.     

双特征空间的实体排序学习

随着大规模知识图谱的出现以及企业高效管理领域知识图谱的需求,知识图谱中的自组织实体检索成为研究热点。给定知识图谱以及用户查询,实体检索的目标在于从给定的知识图谱中返回实体的排序列表。从匹配的角度来看,传统的实体检索模型大都将用户查询和实体统一映射到词的特征空间。这样做具有明显的缺点,例如,将同属于一个实体的两个词视为独立的。为此,该文提出将用户查询和实体同时映射到实体与词两个特征空间方法,称为双特征空间的排序学习。首先将实体抽象成若干个域。之后从词空间和实体空间两个维度分别抽取排序特征,最终应用于排序学习算法中。实验结果表明,在标准数据集上,双特征空间的实体排序学习模型性能显著优于当前先进的实体检索模型。     

Crawling ranked deep Web data sources

In the era of big data, the vast majority of the data are not from the surface Web, the Web that is interconnected by hyperlinks and indexed by most general purpose search engines. Instead, the trove of valuable data often reside in the deep Web, the Web that is hidden behind query interfaces. Since numerous applications, like data integration and vertical portals, require deep Web data, various crawling methods were developed for exhaustively harvesting a deep Web data source with the minimal (or near-minimal) cost. Most existing crawling methods assume that all the documents matched by queries are returned. In practice, data sources often return the top k matches. This makes exhaustive data harvesting difficult: highly ranked documents will be returned multiple times, while documents ranked low have small chance being returned. In this paper, we decompose this problem into two orthogonal sub-problems, i.e., query and ranking bias problems, and propose a document frequency based crawling method to overcome the ranking bias problem. The rational of our method is to use the queries whose document frequencies are within the specified range to avoid the effect of search ranking plus return limit and significantly reduce the difficulty of crawling ranked data source. The method is extensively tested on a variety of datasets and compared with two existing methods. The experimental result demonstrates that our method outperforms the two algorithms by 58 % and 90 % on average respectively.     

基于查询性能预测的鲁棒检索排序研究

信息检索技术致力于从海量的信息资源中为用户获取所需的信息。相较于传统的简单模型,近些年来的大量研究工作在提升了检索结果平均质量的同时,往往忽略了鲁棒性的问题,即造成了很多查询的性能下降,导致用户满意度的显著下降。本文提出了一种基于排序学习的查询性能预测方法,针对每一个查询,对多种模型得到的检索结果列表进行预测,将其中预测性能最优的检索结果列表展示给用户。在LETOR的三个标准数据集OHSUMED、MQ2008和MSLR-WEB10K上的一系列对比实验表明,在以经典的BM25模型作为基准的情况下,与当前最好的检索模型之一LambdaMART相比,该方法在提升了检索结果平均质量的同时,显著地减少了性能下降的查询的数量,具备较好的鲁棒性。