A Novel Anatomy Approach For Document Summarization
Authors: Shahnaz Khadeeja M., Afsar P
Abstract
The phenomenal growth in the number of documents posted on the Internet provides an abundant source of information as an alternative to traditional media. While current technologies are efficient in searching for appropriate documents to satisfy keyword search requests, users still have difficulty assimilating needed knowledge from the over- whelming number of documents. Hence an efficient summarization technique will be needed .In the proposed model we define task called topic anatomy, which summarizes and associates the core parts of topic temporarily so that readers can understand the content easily. The proposed topic anatomy model, called TSCAN, derives the major themes of a topic from the eigenvectors of a temporal block association matrix. Finally, the extracted events are associated through their temporal closeness and context similarity to form the evolution graph of the topic and then we can save our summary into a report format.
An Outlier Detection using K-median Clustering Algorithm
Authors: B. Angelin, Dr. D. Devakumari
Abstract
Clustering is the task of assigning a set of data objects into groups called clusters so that the objects in the same cluster are more similar in some sense to each other than to those in other cluster. Data items whose values are different from rest of the data or whose values fall outside the described range are called outliers. Outlier detection is an important issue in data mining, where it is used to identify and eliminate anomalous data objects from given data set. This research initially reviewed more logics on clustering techniques and outlier detection techniques. Secondly, the problem is identified in K means clustering algorithm for outlier detection. Finally, the solution is proposed from k-median based clustering and compared it with the traditional weight based K- means clustering algorithm. To test the algorithm in outlier detection, the Iris dataset is considered. All the above process are carried out in MATLAB simulations. From the experimental result it is concluded that the proposed method named as K- median weight based approach has the maximum weight value in sepal frame.
Annotating Search Results on Web Database: A Recap
Authors: Malothu Raj Kumar, Dr. S. Jayanthi
Abstract
Nowadays, an increasing number of databases have progressed toward becoming web which are available through HTML shape based hunt interfaces. The information units come back from the fundamental database are generally encoded into the outcome pages progressively for human perusing process. For the encoded information units to be machine handle capable, which is basic for some applications, for example, profound web information accumulation and Internet correlation shopping, they should be extricated out and relegated significant names. In this paper, we display a programmed explanation approach that initially adjusts the information units on an outcome page into various gatherings with the end goal that the information in a similar gathering has the same semantic. At that point, for each gathering we explain it from various viewpoints and total the diverse comments to anticipate a last comment mark for it. An explanation wrapper for the hunt webpage is naturally developed and can be utilized to explain new outcome pages from a similar web database. Our analyses demonstrate that the proposed approach is exceedingly powerful.
An Irreversible Transition towards Cross-layering in Mobile Ad-hoc Network for Multimedia Transmission
Authors: Thangavel. P, Mohanasundaram. S
Abstract
As portable mobile devices gain popularity, the need for Mobile ad-hoc network (MANET) is becoming unavoidable in the wireless-dominated next-generation communication system. MANETs with conventional layered architecture does not efficiently support real-time media transmission since it has large resource requirements and hard timing constraints for data delivery.The cross-layer designs (XLD) differ from a conventional layered architecture where each layer of the architecture performs autonomously. The wide range of different intelligent cross-layer applications reveals its importance. As innovative wireless techniques are implemented, to boost the efficiency of the network, XLDs would be mandatory. On the other hand, there are also some open research issues restraining the growth of cross- layer designs in MANETs. This paper focused on several more radical cross-layer designs that advance significantly from the layered architecture. An illustrative review of existing research on XLDs is presented here. Implementation of XLDs should not affect the accuracy, performance, and efficiency of the current protocol stack. Most of the proposed approaches for the XLD influence the performance of the network considerably. Existing XLDs do not address all the design goals. Prevailing research works on XLD are mostly emphasized on combined solutions involving at most two layers in the protocol stack.