Search

Text summarization solves the problem of presenting the information needed by a user in a compact form. There are different approaches to creating well-formed summaries. One of the newest methods is the Latent Semantic Analysis (LSA). In this paper, different LSA-based summarization algorithms are explained, two of which are proposed by the authors of this paper. The algorithms are evaluated on Turkish and English documents, and their performances are compared using their ROUGE scores. One of our algorithms produces the best scores and both algorithms perform equally well on Turkish and English document sets. [ABSTRACT FROM PUBLISHER]

The information explosion on the Internet has placed high demands on search engines. Despite the improvements in search engine technology, the precision of current search engines is still unsatisfactory. Moreover, the queries submitted by users are short, ambiguous and imprecise. This leads to a number of problems in dealing with similar queries. The problems include lack of common keywords, selection of different documents by the search engine and lack of common clicks etc. These problems render the traditional query clustering methods unsuitable for query recommendations. In this paper, we propose a new query recommendation system. For this, we have identified conceptually related queries by capturing users’ preferences using click-through graphs of web search logs and by extracting the best features, relevant to the queries, from the snippets. The proposed system has an online feature extraction phase and an offline phase in which feature filtering and query clustering are performed. Query clustering is carried out by a new tripartite agglomerative clustering algorithm, Query-Document-Concept Clustering, in which the documents are used innovatively to decouple queries and features/concepts in a tripartite graph structure. This results in clusters of similar queries, associated clusters of documents and clusters of features. We model the query recommendation problem in four different ways. Two models are non-personalized and personalized content-ignorant models. Other two are non-personalized and personalized content-aware models. Three similarity measures are introduced to estimate different kinds of similarities. Experimental results show that the proposed approach has better precision, recall and F-measure than the existing approaches. [ABSTRACT FROM PUBLISHER]

Blogs are one of the main user-generated contents on the web and are growing in number rapidly. The characteristics of blogs require the development of specialized search methods which are tuned for the blogosphere. In this paper, we focus on blog retrieval, which aims at ranking blogs with respect to their recurrent relevance to a user’s topic. Although different blog retrieval algorithms have already been proposed, few of them have considered temporal properties of the input queries. Therefore, we propose an efficient approach to improving relevant blog retrieval using temporal property of queries. First, time sensitivity of each query is automatically computed for different time intervals based on an initially retrieved set of relevant posts. Then a temporal score is calculated for each blog and finally all blogs are ranked based on their temporal and content relevancy with regard to the input query. Experimental analysis and comparison of the proposed method are carried out using a standard dataset with 45 diverse queries. Our experimental results demonstrate that, using different measurement criteria, our proposed method outperforms other blog retrieval methods. [ABSTRACT FROM AUTHOR]

Mobile phones have been identified as one of the technologies that can be used to overcome the challenges of information dissemination regarding serious diseases. Short message services, a much used function of cell phones, for example, can be turned into a major tool for accessing databases. This paper focuses on the design and development of a short message services-based information access algorithm to carefully screen information on human immunodeficiency virus/acquired immune deficiency syndrome within the context of a frequently asked questions system. However, automating the short message services-based information search and retrieval poses significant challenges because of the inherent noise in its communications. The developed algorithm was used to retrieve the best-ranked question–answer pair. Results were evaluated using three metrics: average precision, recall and computational time. The retrieval efficacy was measured and it was confirmed that there was a significant improvement in the results of the proposed algorithm when compared with similar retrieval algorithms. [ABSTRACT FROM PUBLISHER]

In this paper, we are interested in XML multimedia retrieval, the aim of which is to find relevant multimedia objects such as images, audio and video through their context as document structure. In context-based multimedia retrieval, the most common technique is based on the text surrounding the image. However, such textual information can be irrelevant to the image content. Therefore many works are oriented to the use of alternative techniques to extend the image description, such as the use of ontologies, relevance feedback, and user profiles. We studied in our work the use of links between XML elements to improve image retrieval. More precisely, we propose dividing the document into regions through the document structure and image position. Then we weight links between these regions according to their hierarchical positions, in order to distinguish between links that are useful and those that are not useful. We then apply an updated version of the HITS algorithm at the region level, and compute a final image score by combining link scores with initial image scores. Experiments were done on the INEX 2006 and 2007 multimedia tracks, and showed the potential of our method. [ABSTRACT FROM PUBLISHER]

Association Rule Mining (ARM) is one of the data mining techniques used to extract hidden knowledge from datasets, that can be used by an organization's decision makers to improve overall profit. However, performing ARM requires repeated passes over the entire database. Obviously, for large database, the role of input/output overhead in scanning the database is very significant. A popular solution to improve the speed of ARM is to apply the mining algorithm on a sample instead of the entire database. In this paper, a parameterized sampling algorithm for ARM is presented. This algorithm extracts sample datasets based on three parameters: transaction frequency, transaction length and transaction frequency-length. To evaluate its performance and accuracy, a comparison against a two-phase sampling-based algorithm is performed using real and synthetic datasets. The experimental results show that the proposed sampling algorithm in some cases outperforms two-phase sampling algorithm, and achieves up to 98% accuracy. [ABSTRACT FROM AUTHOR]

On the Web, hyperlinks have been used both to assess the impact of academic Web sites and to trace aspects of online informal scholarly communication. They are also used in Web information retrieval algorithms to identify important pages and to cluster pages by topic, both of which help in ranking search engine results. In this paper we investigate a type of link that is of particular interest for all of these applications: one that crosses subject boundaries. We took a sample of 586 linked pairs of domains in different UK academic sites, and extracted those that represented different subjects, resulting in 52 pairs of domains with different subjects. These were then grouped by the type of relationship between the source and target page. Over a third of the links formed a scholarly connection between similar subjects, but in 8% of cases dissimilar subjects also had a scholarly connection. Additionally, higher education teaching links were seen to form an extensive cross- disciplinary network, accounting for 19% of the links. A significant number of links (12%) also targeted non-subject- specific general resources. The results suggest that mapping disciplinary collaboration on the Web should be feasible but that this process and topic identification in academic Webs would both be helped by the prior removal of key higher education teaching and popular general pages from the data set. These, and computing pages to a lesser extent, play a role more pernicious than `stop words' in traditional information retrieval. The conclusions are of a qualitative rather than quantitative nature because of the small effective sample size, so an initial set of thousands of links would be required to remedy this. [ABSTRACT FROM AUTHOR]

Ontology fusion in knowledge bases has become less easy, due to the massive capacity involved in the process of semantic similarity calculation. Many similarity calculation methods have been developed, although they are hardly united. This article contributes a low-cost similarity calculation method for ontology fusion, based on the inspiration of binary metrics, with the aim of reducing the size of similarity calculations both spatially and logically. By introducing the definitions of a heterogeneous ontology, entities of ontologies and rules of ontology fusion on the basis of concept fusion and relationship fusion, we put forward the algorithm of main traverse procedure and calculated to be the least cost in time and space in comparison with traditional methods. We adopted three experiments to testify the usability of our approach from the perspective of actual library resources, small datasets and large datasets. In Experiment 1, the bibliographic data from East China Normal University Library were used to show the feasibility and capability of our proposal and present the process of the algorithm. In both Experiments 2 and 3, our approach had at least 88% confidence in detecting accurate merging mappings and also decreased time cost. The test demonstrated a good fusion result. The problem of lower recalls caused by error analysis results from the conflict between the complex structures in ontologies and the recursive functions, which will be improved in the future. [ABSTRACT FROM AUTHOR]

Spatial data indexing methods are of extreme importance as they massively build up as a result of the explosive growth in capturing data with spatial features. No matter how much the data size is, eventually it will reside on disk pages. Disk pages have to be properly indexed to preserve spatial properties of objects, optimise disk space usage and improve objects' retrieval performance. One of the most popular spatial data indexes is the R-tree which is a height balanced tree data structure, where leaf nodes resemble disk pages and contain pointers to objects' locations. A single tree node can host up to a maximum number of objects, where any more insertion makes it an overflown node and it has to be split. Better splits lead to better index performance and more utilisation of disk space. In this work, we introduce a new way of finding the most proper split for an overflown node in the R-tree index. The proposed work scans – in a linear cost – the overflown node's objects once to identify the distribution of objects' locations (minimum bounding rectangles (MBRs)) in relative to its node's bounding rectangle (node's MBR). It uses objects' locations to calculate – for each main axis – the split quality factors: expected overlap between resulting nodes, objects distribution evenness among resulting nodes and the perimeter of resulting nodes. The axis with better combined quality factors values is selected as the split axis. The Splitting based on Objects' Locations Distribution (SOLD) algorithm was implemented and tested against two other splitting algorithms, experiments using synthetic and real data files showed good results and it outperformed both algorithms in index creation tests and data retrieval tests. [ABSTRACT FROM AUTHOR]

Ontology matching is an essential aspect of the Semantic Web with a goal of finding alignments among the entities of given ontologies. Ontology matching is a necessary step for establishing interoperation and knowledge sharing among Semantic Web applications. In this study we present an algorithm and a tool developed based on this algorithm to find correspondences among entities of input ontologies. The proposed algorithm uses a new lexical similarity measure and also utilizes structural information of ontologies to determine their corresponding entities. The lexical similarity measure generates a bag of words for each entity based on its label and description information. The structural approach creates a grid for each node in the ontologies. The combination of lexical and structural approaches creates the similarity matrix between the source and target ontologies. The proposed algorithm was tested on a well known benchmark and also compared to other algorithms presented in the literature. Our experimental results show the proposed algorithm is effective and outperforms other algorithms. [ABSTRACT FROM AUTHOR]

In this study we investigated the stemming mechanisms of Google. We used its web interface and submitted many queries via a program. Stemming is the process of correlating morphologically similar words with one another. Search engines use stemming to match documents having one form of a word with queries having another form of the same word. We investigated the stemming mechanism of Google for three classes of words: singulars/plurals, combined words, and verbs with many postfixes. Our results indicate that Google uses a document-based algorithm for stemming. It evaluates each document separately and makes a decision to index or not for the conflated forms of the words it has. It indexes documents only for word forms that are semantically strongly correlated. While it indexes documents for singulars and plurals frequently, it rarely indexes documents for word forms with the postfixes of -able or -tively. [ABSTRACT FROM AUTHOR]

We present a unified treatment of a number of related in-place MSD radix sort algorithms with varying radices, collectively referred to here as ‘Matesort’ algorithms. These algorithms use the idea of in-place partitioning which is a considerable improvement over the traditional linked list implementation of radix sort that uses O(n) space. The binary Matesort algorithm is a recast of the classical radix-exchange sort, emphasizing the role of in-place partitioning and efficient implementation of bit processing operations. This algorithm is O(k) space and has O(kn) worst-case order of running time, where k is the number of bits needed to encode an element value and n is the number of elements to be sorted. The binary Matesort algorithm is evolved into a number of other algorithms including ‘continuous Matesort’ for handling floating point numbers, and a number of ‘general radix Matesort’ algorithms. We present formulation and analysis for three different approaches (sequential, divide-and-conquer and permutation-loop) for partitioning by the general radix Matesort algorithm. The divide-and-conquer approach leads to an elegantly coded algorithm with better performance than the permutation-loop-based American Flag Sort algorithm. [ABSTRACT FROM AUTHOR]