Document Retrieval

Detecting academic plagiarism is a pressing problem, e.g., for educational and research institutions, funding agencies, and academic publishers. Existing plagiarism detection systems reliably identify copied text, or near copies of text, but often fail to detect disguised forms of academic plagiarism, such as paraphrases, translations, and idea plagiarism. We present Semantic Concept Pattern Analysis - an approach that performs an integrated analysis of semantic text relatedness and structural text similarity. Using 25 officially retracted academic plagiarism cases, we demonstrate that our approach can detect plagiarism that established text matching approaches would not identify. We view our approach as a promising addition to improve the detection capabilities for strong paraphrases. We plan to further improve Semantic Concept Pattern Analysis and include the approach as part of an integrated detection process that analyzes heterogeneous similarity features to better identify the many possible forms of plagiarism in academic documents.

In this paper we introduce COV, a novel information retrieval (IR) algorithm for massive databases based on vector space modeling and spectral analysis of the covariance matrix, for the document vectors, to reduce the scale of the problem. Since the dimension of the covariance matrix depends on the attribute space and is independent of the number of documents, COV can be applied to databases that are too massive for methods based on the singular value decomposition of the document-attribute matrix, such as latent semantic indexing (LSI). In addition to improved scalability, theoretical considerations indicate that results from our algorithm tend to be more accurate than those from LSI, particularly in detecting subtle differences in document vectors. We demonstrate the power and accuracy of COV through an important topic in data mining, known as outlier cluster detection. We propose two new algorithms for detecting major and outlier clusters in databases—the first is based on LSI, and the second on COV. Our implementation studies indicate that our cluster detection algorithms outperform the basic LSI and COV algorithm in detecting outlier clusters.

A 24-dimensional model for the ‘harmonic content’ of pieces of music has proved to be remarkably robust in the retrieval of polyphonic queries from a database of polyphonic music in the presence of quite significant noise and errors in either query or database document. We have further found that higher-order (1st- to 3rd-order) models tend to work better for music retrieval than 0th-order ones owing to the richer context they capture. However, there is a serious performance cost due to the large size of such models and the present paper reports on some attempts to reduce dimensionality while retaining the general robustness of the method. We find that some simple reduced-dimensionality models, if their parameter settings are carefully chosen, do indeed perform almost as well as the full 24-dimensional versions. Furthermore, in terms of recall in the top 1000 documents retrieved, we find that a 6-dimensional 2nd-order model gives even better performance than the full model. This represents a potential 64-times reduction in model size and search-time, making it a suitable candidate for filtering a large database as the first stage of a two-stage retrieval system.

The motivation for this research is the emergence of mobile information systems where information is disseminated to mobile individuals via handheld devices. A key distinction between mobile and desktop computing is the significance of the relationship between the spatial location of an individual and the spatial location associated with information accessed by that individual. Given a set of spatially referenced documents retrieved from a mobile information system, this set can be presented using alternative interfaces of which two presently dominate: textual lists and graphical two-dimensional maps. The purpose of this paper is to explore how mixed reality interfaces can be used for the presentation of information on mobile devices.

Multimedia content analysis refers to the computerized understanding of the semantic meanings of a multimedia document, such as a video sequence with an accompanying audio track. With a multimedia document, its semantics are embedded in multiple forms that are usually complimentary of each other, Therefore, it is necessary to analyze all types of data: image frames, sound tracks, texts that can be extracted from image frames, and spoken words that can be deciphered from the audio track. This usually involves segmenting the document into semantically meaningful units, classifying each unit into a predefined scene type, and indexing and summarizing the document for efficient retrieval and browsing. We review advances in using audio and visual information jointly for accomplishing the above tasks. We describe audio and visual features that can effectively characterize scene content, present selected algorithms for segmentation and classification, and review some testbed systems for video archiving and retrieval. We also describe audio and visual descriptors and description schemes that are being considered by the MPEG-7 standard for multimedia content description

We are developing a web-based plagiarism detection system to detect plagiarism in written Arabic documents. This paper describes the proposed framework of our plagiarism detection system. The proposed plagiarism detection framework comprises of two main components, one global and the other local. The global component is heuristics-based, in which a potentially plagiarized given document is used to construct a set of representative queries by using different best performing heuristics. These queries are then submitted to Google via Google's search API to retrieve candidate source documents from the Web. The local component carries out detailed similarity computations by combining different similarity computation techniques to check which parts of the given document are plagiarised and from which source documents retrieved from the Web. Since this is an ongoing research project, the quality of overall system is not evaluated yet.

This paper presents a method to improve the performance of Information Retrieval System (IRS) by increasing the no of relevant documents retrieved. There are several types of uncertainty and fuzziness associated with IRS like search term uncertainty, relevance uncertainty involved in retrieving of irrelevant documents. The aim of this paper is to eliminate different types of uncertainty and increase the chance of retrieving relevant documents. In the framework a method is proposed which first calculate query and document cluster similarity which not only retrieve the documents matching query terms as well as similar to retrieved documents by calculating the query and cluster similarity. This helps to reduce search term uncertainty and tries to reduce the fuzziness associated with document relevance in two steps. First modification is made in general term frequency-inverse document frequency (tf-idf) scoring mechanism to give importance of informativeness of a document contents and secondly calculating query and document summary overlap. All the above information is used to measure the document relevant score. Finally retrieved documents are filtered by Pearson correlation coefficient between query vector and document vector to find out only those documents correlated with query. In experiment standard NPL test collection prepared by Vaswani and Cameron at the National Physical Laboratory in England was used. After full implementation of above methodology it was found that proposed work is better in comparison with existing methods.