A scaleable document clustering approach for large document corpora

In this paper, the scalability and quality of the contextual document clustering (CDC) approach is demonstrated for large data-sets using the whole Reuters Corpus Volume 1 (RCV1) collection. CDC is a form of distributional clustering, which automatically discovers contexts of narrow scope within a document corpus. These contexts act as attractors for clustering documents that are semantically related to each other. Once clustered, the documents are organized into a minimum spanning tree so that the topical similarity of adjacent documents within this structure can be assessed. The pre-defined categories from three different document category sets are used to assess the quality of CDC in terms of its ability to group and structure semantically related documents given the contexts. Quality is evaluated based on two factors, the category overlap between adjacent documents within a cluster, and how well a representative document categorizes all the other documents within a cluster. As the RCV1 collection was collated in a time ordered fashion, it was possible to assess the stability of clusters formed from documents within one time interval when presented with new unseen documents at subsequent time intervals. We demonstrate that CDC is a powerful and scaleable technique with the ability to create stable clusters of high quality. Additionally, to our knowledge this is the first time that a collection as large as RCV1 has been analyzed in its entirety using a static clustering approach.     

Mining fuzzy frequent itemsets for hierarchical document clustering

As text documents are explosively increasing in the Internet, the process of hierarchical document clustering has been proven to be useful for grouping similar documents for versatile applications. However, most document clustering methods still suffer from challenges in dealing with the problems of high dimensionality, scalability, accuracy, and meaningful cluster labels. In this paper, we will present an effective Fuzzy Frequent Itemset-Based Hierarchical Clustering (F²IHC) approach, which uses fuzzy association rule mining algorithm to improve the clustering accuracy of Frequent Itemset-Based Hierarchical Clustering (FIHC) method. In our approach, the key terms will be extracted from the document set, and each document is pre-processed into the designated representation for the following mining process. Then, a fuzzy association rule mining algorithm for text is employed to discover a set of highly-related fuzzy frequent itemsets, which contain key terms to be regarded as the labels of the candidate clusters. Finally, these documents will be clustered into a hierarchical cluster tree by referring to these candidate clusters. We have conducted experiments to evaluate the performance based on Classic4, Hitech, Re0, Reuters, and Wap datasets. The experimental results show that our approach not only absolutely retains the merits of FIHC, but also improves the accuracy quality of FIHC.     

Quantifying the impact of concept recognition on biomedical information retrieval

In ad hoc querying of document collections, current approaches to ranking primarily rely on identifying the documents that contain the query terms. Methods such as query expansion, based on thesaural information or automatic feedback, are used to add further terms, and can yield significant though usually small gains in effectiveness. Another approach to adding terms, which we investigate in this paper, is to use natural language technology to annotate - and thus disambiguate - key terms by the concept they represent. Using biomedical research documents, we quantify the potential benefits of tagging users’ targeted concepts in queries and documents in domain-specific information retrieval. Our experiments, based on the TREC Genomics track data, both on passage and full-text retrieval, found no evidence that automatic concept recognition in general is of significant value for this task. Moreover, the issues raised by these results suggest that it is difficult for such disambiguation to be effective.     

Re-ranking model based on document clusters

In this paper, we describe a model of information retrieval system that is based on a document re-ranking method using document clusters. In the first step, we retrieve documents based on the inverted-file method. Next, we analyze the retrieved documents using document clusters, and re-rank them. In this step, we use static clusters and dynamic cluster view. Consequently, we can produce clusters that are tailored to characteristics of the query. We focus on the merits of the inverted-file method and cluster analysis. In other words, we retrieve documents based on the inverted-file method and analyze all terms in document based on the cluster analysis. By these two steps, we can get the retrieved results which are made by the consideration of the context of all terms in a document as well as query terms. We will show that our method achieves significant improvements over the method based on similarity search ranking alone.     

Technology survey and invalidity search: A comparative study of different tasks for Japanese patent document retrieval

A comparative study of two types of patent retrieval tasks, technology survey and invalidity search, using the NTCIR-3 and -4 test collections is described, with a focus on pseudo-feedback effectiveness and different retrieval models. Invalidity searches are peculiar to patent retrieval tasks and feature small numbers of relevant documents and long queries. Different behaviors of effectiveness are observed when applying different retrieval models and pseudo-feedback. These different behaviors are analyzed in terms of the “weak cluster hypothesis”, i.e., terminological cohesiveness through relevant documents.     

Optimization and label propagation in bipartite heterogeneous networks to improve transductive classification of texts

Transductive classification is a useful way to classify texts when labeled training examples are insufficient. Several algorithms to perform transductive classification considering text collections represented in a vector space model have been proposed. However, the use of these algorithms is unfeasible in practical applications due to the independence assumption among instances or terms and the drawbacks of these algorithms. Network-based algorithms come up to avoid the drawbacks of the algorithms based on vector space model and to improve transductive classification. Networks are mostly used for label propagation, in which some labeled objects propagate their labels to other objects through the network connections. Bipartite networks are useful to represent text collections as networks and perform label propagation. The generation of this type of network avoids requirements such as collections with hyperlinks or citations, computation of similarities among all texts in the collection, as well as the setup of a number of parameters. In a bipartite heterogeneous network, objects correspond to documents and terms, and the connections are given by the occurrences of terms in documents. The label propagation is performed from documents to terms and then from terms to documents iteratively. Nevertheless, instead of using terms just as means of label propagation, in this article we propose the use of the bipartite network structure to define the relevance scores of terms for classes through an optimization process and then propagate these relevance scores to define labels for unlabeled documents. The new document labels are used to redefine the relevance scores of terms which consequently redefine the labels of unlabeled documents in an iterative process. We demonstrated that the proposed approach surpasses the algorithms for transductive classification based on vector space model or networks. Moreover, we demonstrated that the proposed algorithm effectively makes use of unlabeled documents to improve classification and it is faster than other transductive algorithms.     

Word sense discrimination in information retrieval: A spectral clustering-based approach

Word sense ambiguity has been identified as a cause of poor precision in information retrieval (IR) systems. Word sense disambiguation and discrimination methods have been defined to help systems choose which documents should be retrieved in relation to an ambiguous query. However, the only approaches that show a genuine benefit for word sense discrimination or disambiguation in IR are generally supervised ones. In this paper we propose a new unsupervised method that uses word sense discrimination in IR. The method we develop is based on spectral clustering and reorders an initially retrieved document list by boosting documents that are semantically similar to the target query. For several TREC ad hoc collections we show that our method is useful in the case of queries which contain ambiguous terms. We are interested in improving the level of precision after 5, 10 and 30 retrieved documents (P@5, P@10, P@30) respectively. We show that precision can be improved by 8% above current state-of-the-art baselines. We also focus on poor performing queries.     

A query term re-weighting approach using document similarity

Pseudo-relevance feedback is the basis of a category of automatic query modification techniques. Pseudo-relevance feedback methods assume the initial retrieved set of documents to be relevant. Then they use these documents to extract more relevant terms for the query or just re-weigh the user's original query. In this paper, we propose a straightforward, yet effective use of pseudo-relevance feedback method in detecting more informative query terms and re-weighting them. The query-by-query analysis of our results indicates that our method is capable of identifying the most important keywords even in short queries. Our main idea is that some of the top documents may contain a closer context to the user's information need than the others. Therefore, re-examining the similarity of those top documents and weighting this set based on their context could help in identifying and re-weighting informative query terms. Our experimental results in standard English and Persian test collections show that our method improves retrieval performance, in terms of MAP criterion, up to 7% over traditional query term re-weighting methods.     

Adapting information retrieval to query contexts

In current IR approaches documents are retrieved only according to the terms specified in the query. The same answers are returned for the same query whatever the user and the search goal are. In reality, many other contextual factors strongly influence document’s relevance and they should be taken into account in IR operations. This paper proposes a method, based on language modeling, to integrate several contextual factors so that document ranking will be adapted to the specific query contexts. We will consider three contextual factors in this paper: the topic domain of the query, the characteristics of the document collection, as well as context words within the query. Each contextual factor is used to generate a new query language model to specify some aspect of the information need. All these query models are then combined together to produce a more complete model for the underlying information need. Our experiments on TREC collections show that each contextual factor can positively influence the IR effectiveness and the combined model results in the highest effectiveness. This study shows that it is both beneficial and feasible to integrate more contextual factors in the current IR practice.     

不同被引频次论文主题词组合特征及其与论文新颖性关系的研究——以免疫学ESI指标为例

[目的/意义]探索论文被引次数是否和论文内容即概念组合方式有关。[方法/过程]选取WoS数据库中的免疫学科,抽取其中高、中、低被引频次三种论文集合的主题词,分析各集合主题词频次分布的集中离散趋势。分别构建主题词共现网络,通过网络拓扑属性的分析,了解三种论文集合在概念组合方式上的异同,衡量非典型组合与新颖性的关系。[结果/结论](1)不同被引频次的文献集合在主题类型的分布和主题词分散程度上有较大差异。(2)高被引和中被引论文集的主题词共现网络具有小世界性,低被引论文集的主题词网络不具有小世界性。(3)高被引论文集的主题词共现网络比较紧密,且主题词非典型组合的比例要高于其他两种论文集。低被引论文集的主题词网络比较松散。论文的被引次数与其主题热度、主题之间联系密切程度以及主题之间组合方式相关。     

分布式检索中查询结果合并策略研究

查询结果合并是分布式信息检索的重要步骤。本文依据选中信息集中文档重叠的程度以及信息集的同构、异构性，将查询结果的合并策略分3种情况进行分析：选中的信息集所含文档没有或有少量的重叠，选中的信息集同构，选中的信息集异构且所含文档有部分重叠。指出查询结果合并策略的深入研究，对于促进分布式检索技术的发展具有积极意义。     

Parsimonious translation models for information retrieval

In the KL divergence framework, the extended language modeling approach has a critical problem of estimating a query model, which is the probabilistic model that encodes the user’s information need. For query expansion in initial retrieval, the translation model had been proposed to involve term co-occurrence statistics. However, the translation model was difficult to apply, because the term co-occurrence statistics must be constructed in the offline time. Especially in a large collection, constructing such a large matrix of term co-occurrences statistics prohibitively increases time and space complexity. In addition, reliable retrieval performance cannot be guaranteed because the translation model may comprise noisy non-topical terms in documents. To resolve these problems, this paper investigates an effective method to construct co-occurrence statistics and eliminate noisy terms by employing a parsimonious translation model. The parsimonious translation model is a compact version of a translation model that can reduce the number of terms containing non-zero probabilities by eliminating non-topical terms in documents. Through experimentation on seven different test collections, we show that the query model estimated from the parsimonious translation model significantly outperforms not only the baseline language modeling, but also the non-parsimonious models.     

A results merging algorithm for distributed information retrieval environments that combines regression methodologies with a selective download phase

The problem of results merging in distributed information retrieval environments has gained significant attention the last years. Two generic approaches have been introduced in research. The first approach aims at estimating the relevance of the documents returned from the remote collections through ad hoc methodologies (such as weighted score merging, regression etc.) while the other is based on downloading all the documents locally, completely or partially, in order to calculate their relevance. Both approaches have advantages and disadvantages. Download methodologies are more effective but they pose a significant overhead on the process in terms of time and bandwidth. Approaches that rely solely on estimation on the other hand, usually depend on document relevance scores being reported by the remote collections in order to achieve maximum performance. In addition to that, regression algorithms, which have proved to be more effective than weighted scores merging algorithms, need a significant number of overlap documents in order to function effectively, practically requiring multiple interactions with the remote collections. The new algorithm that is introduced is based on adaptively downloading a limited, selected number of documents from the remote collections and estimating the relevance of the rest through regression methodologies. Thus it reconciles the above two approaches, combining their strengths, while minimizing their drawbacks, achieving the limited time and bandwidth overhead of the estimation approaches and the increased effectiveness of the download. The proposed algorithm is tested in a variety of settings and its performance is found to be significantly better than the former, while approximating that of the latter.     

Estimating Gaussian mixture models in the local neighbourhood of embedded word vectors for query performance prediction

The study of query performance prediction (QPP) in information retrieval (IR) aims to predict retrieval effectiveness. The specificity of the underlying information need of a query often determines how effectively can a search engine retrieve relevant documents at top ranks. The presence of ambiguous terms makes a query less specific to the sought information need, which in turn may degrade IR effectiveness. In this paper, we propose a novel word embedding based pre-retrieval feature which measures the ambiguity of each query term by estimating how many ‘senses’ each word is associated with. Assuming each sense roughly corresponds to a Gaussian mixture component, our proposed generative model first estimates a Gaussian mixture model (GMM) from the word vectors that are most similar to the given query terms. We then use the posterior probabilities of generating the query terms themselves from this estimated GMM in order to quantify the ambiguity of the query. Previous studies have shown that post-retrieval QPP approaches often outperform pre-retrieval ones because they use additional information from the top ranked documents. To achieve the best of both worlds, we formalize a linear combination of our proposed GMM based pre-retrieval predictor with NQC, a state-of-the-art post-retrieval QPP. Our experiments on the TREC benchmark news and web collections demonstrate that our proposed hybrid QPP approach (in linear combination with NQC) significantly outperforms a range of other existing pre-retrieval approaches in combination with NQC used as baselines.     

A general matrix framework for modelling Information Retrieval

In this paper, we present a well-defined general matrix framework for modelling Information Retrieval (IR). In this framework, collections, documents and queries correspond to matrix spaces. Retrieval aspects, such as content, structure and semantics, are expressed by matrices defined in these spaces and by matrix operations applied on them. The dualities of these spaces are identified through the application of frequency-based operations on the proposed matrices and through the investigation of the meaning of their eigenvectors. This allows term weighting concepts used for content-based retrieval, such as term frequency and inverse document frequency, to translate directly to concepts for structure-based retrieval. In addition, concepts such as pagerank, authorities and hubs, determined by exploiting the structural relationships between linked documents, can be defined with respect to the semantic relationships between terms. Moreover, this mathematical framework can be used to express classical and alternative evaluation measures, involving, for instance, the structure of documents, and to further explain and relate IR models and theory. The high level of reusability and abstraction of the framework leads to a logical layer for IR that makes system design and construction significantly more efficient, and thus, better and increasingly personalised systems can be built at lower costs.     

A link-based collection fusion strategy

This paper presents a method for solving the collection fusion problem in hypermedia digital libraries. The proposition which is explored and evaluated is that across document links between hypermedia documents residing in distributed hypermedia collections can supply sufficient useful information to allow effective collection fusion. In contrast to other collection fusion strategies, the link-based fusion strategy does not require a learning phase before it can be utilised and, also does not use any information from remote collections other than the returned list of documents. Because of these characteristics the proposed fusion strategy is suitable for very large and extremely dynamic environments in which other collection fusion strategies (e.g. learning collection fusion strategies) may be inapplicable. Evaluation of the link-based fusion strategy demonstrates that the proposed strategy is more effective and efficient than the uniform strategy which can be applied under the same conditions.     

Concept unification of terms in different languages via web mining for Information Retrieval

For historical and cultural reasons, English phases, especially proper nouns and new words, frequently appear in Web pages written primarily in East Asian languages such as Chinese, Korean, and Japanese. Although such English terms and their equivalences in these East Asian languages refer to the same concept, they are often erroneously treated as independent index units in traditional Information Retrieval (IR). This paper describes the degree to which the problem arises in IR and proposes a novel technique to solve it. Our method first extracts English terms from native Web documents in an East Asian language, and then unifies the extracted terms and their equivalences in the native language as one index unit. For Cross-Language Information Retrieval (CLIR), one of the major hindrances to achieving retrieval performance at the level of Mono-Lingual Information Retrieval (MLIR) is the translation of terms in search queries which can not be found in a bilingual dictionary. The Web mining approach proposed in this paper for concept unification of terms in different languages can also be applied to solve this well-known challenge in CLIR. Experimental results based on NTCIR and KT-Set test collections show that the high translation precision of our approach greatly improves performance of both Mono-Lingual and Cross-Language Information Retrieval.     

A split-list approach for relevance feedback in information retrieval

In this paper we present a new algorithm for relevance feedback (RF) in information retrieval. Unlike conventional RF algorithms which use the top ranked documents for feedback, our proposed algorithm is a kind of active feedback algorithm which actively chooses documents for the user to judge. The objectives are (a) to increase the number of judged relevant documents and (b) to increase the diversity of judged documents during the RF process. The algorithm uses document-contexts by splitting the retrieval list into sub-lists according to the query term patterns that exist in the top ranked documents. Query term patterns include a single query term, a pair of query terms that occur in a phrase and query terms that occur in proximity. The algorithm is an iterative algorithm which takes one document for feedback in each of the iterations. We experiment with the algorithm using the TREC-6, -7, -8, -2005 and GOV2 data collections and we simulate user feedback using the TREC relevance judgements. From the experimental results, we show that our proposed split-list algorithm is better than the conventional RF algorithm and that our algorithm is more reliable than a similar algorithm using maximal marginal relevance.     

Patent document categorization based on semantic structural information

The number of patent documents is currently rising rapidly worldwide, creating the need for an automatic categorization system to replace time-consuming and labor-intensive manual categorization. Because accurate patent classification is crucial to search for relevant existing patents in a certain field, patent categorization is a very important and useful field. As patent documents are structural documents with their own characteristics distinguished from general documents, these unique traits should be considered in the patent categorization process. In this paper, we categorize Japanese patent documents automatically, focusing on their characteristics: patents are structured by claims, purposes, effects, embodiments of the invention, and so on. We propose a patent document categorization method that uses the k-NN (k-Nearest Neighbour) approach. In order to retrieve similar documents from a training document set, some specific components to denote the so-called semantic elements, such as claim, purpose, and application field, are compared instead of the whole texts. Because those specific components are identified by various user-defined tags, first all of the components are clustered into several semantic elements. Such semantically clustered structural components are the basic features of patent categorization. We can achieve a 74% improvement of categorization performance over a baseline system that does not use the structural information of the patent.