Ni Lar Thein

An Efficient Indexing Mechanism for Data Deduplication

At present, there is a vast amount of duplicated data or redundant data in storage systems. Data de-duplication can eliminate multiple copies of the same file and duplicated segments or chunks of data within those files. In these days, therefore, data de-duplication becomes an interesting field in storage environments especially in persistent data storage for data centers. Many data deduplication mechanisms have been proposed for efficient data deduplication in order to safe storage space. Current issue for data deduplication is to avoid full-chunk indexing to identify the incoming data is new, which is time consuming process. In this paper, we propose an efficient indexing mechanism for this problem using the advantage of B+ tree properties. In our proposed system, we will first separate the file into variable-length chunks using Two Thresholds Two Divisors chunking algorithm. ChunkIDs are then obtained by applying hash function to the chunks. The resulted ChunkIDs are used to build as indexing keys in B+ tree like index structure. So the searching time for the duplicate file chunks reduces from O (n) to O (log n), which can avoid the risk of full chunk indexing.

A Persistent Labeling Scheme for Dynamic Ordered XML Trees

XML is becoming the new standard for the exchange and publishing of data over the Internet. Documents obeying the XML standard can be viewed as trees basically the parse tree of the document. XML database systems often give each item in the document (node in the tree) a unique logical identifier (called a label) and use those labels for an efficient processing of queries, in particular queries involving structural conditions or testing for changes in the document content. Therefore several path indexing and numbering schemes have been proposed. XML data on the Web are subjected to frequent updates. During the update on XML data, most of these approaches can need to recompute existing labels, which is rather time consuming. The goal of our labeling scheme is to design a persistent structural labeling scheme that supports the representation of ancestor-descendant relationship and sibling relationship between nodes. Moreover our labeling scheme supports insertion of new nodes at arbitrary positions in the XML tree without re-labeling and without conflicting existing labels

Mobile Agents Based Load Balancing Method for Parallel Applications

A parallel program is composed of multiple processes, each of which is to perform one or more tasks defined by the program. The optimization objective for partitioning is to balance the workload among processes and to minimize the interprocess communication needs. Its primary goal is to ensure that no processor is idle while there are processes waiting for services in other processors. Mobile agent technology has the ability to travel from host to host in a network and to places of their own choosing at times. It motivates force in reducing force in reducing network traffic and overcoming network latency and their ability to operate asynchronously and independently of the process that created them, mobile agent has been an effective choice for parallel and distributed computing. In this paper, we propose a framework called MALB (mobile agent based load balancing) that uses mobile agent technology to implement efficient load balancing on computing cluster nodes. The MALB system is designed to control the all slave nodes from the master for evenly distributing workloads among processor. The key features of MALB include evaluation parallel tasks on behalf of the user. We experiment MALB with sample applications and found out. Various load balancing policies can be incorporated with versatile mobile agents in the framework. Mobile agents based load balancing approaches have the merits of high flexibility, low network traffic and high asynchrony

Domain Adaptive Information Extraction Using Link Grammar and WordNet

Nowadays, people want to extract variety of information from on line texts. As more and more text becomes available on-line, there is emergent need for systems that extract information automatically from text corpus. One of the principle challenges of information extraction is the efficient customization of a system to a new domain. Adapting an information extraction system to a new domain entails the construction of a new set of extraction rules. Many recent information extraction systems have ignored the tedious and time-consuming nature of that process. This paper proposes an alternative approach, which generate candidate extraction rules from untagged text corpus using Link Grammar Parser and filter the final extraction rules using Wordnet and linguistic patterns. The proposed method not only reduces the amount of time and effort required to create an appropriate training corpus but also obviates the need to examine many candidate extraction rules so that the system can easily port well to different domain.

To Visualize the Coupling among Modules

There are levels of coupling and cohesion between classes in object oriented systems as well as between modules. Low module coupling is one of the quantifying the module coupling. The degree of coupling is a measure of the software quality. So far, several studies have been done on quantifying the module coupling and relating the results to effective software design process, in this article, we propose new some coupling such as import call coupling, import scalar coupling, import stamp coupling, import tramp coupling, export call coupling, export scalar coupling, export tramp coupling, and export stamp coupling and try to develop a software system for visualization of the module coupling. Module coupling visualization can help programmers to see bottlenecks in their design and re-engineer it for improvement

Efficiently Creating Dynamic Web Content: A Fragment Based Approach

Fragment based Web publication can significantly improve performance and increase the cacheability of dynamic Web data. In order for a Web site to use fragment-based content generation, however, good methods are needed for fragmenting the Web pages. Manual fragmentation of Web pages is expensive, error prone, and unscalable. This paper proposes a scheme to automatically detect and flag fragments that are cost-effective cache units in Web sites serving dynamic content. This approach fragments the Web pages with respect to their information sharing behavior, personalization characteristics, and change patterns. This paper proposes a framework for fragment detection, which includes a hierarchical and fragment aware model for dynamic Web pages and a compact and effective data structure for fragment detection This proposed system improves the efficiency of disk-space utilization at edge caches and reduces the loads on the network and the origin server

Translation model of myanmar phrases for statistical machine translation

In this paper, we present a translation model which uses syntactic structure and morphology of Myanmar language to improve Myanmar to English machine translation system. This system is implemented as a subsystem of Myanmar to English translation system and based on statistical approach by using Myanmar-English Bilingual corpus. It also uses two types of information: language model and translation model. The source language model is based on N-gram method to extract phrases from segmented Myanmar sentences and the translation model is based on syntactic structure, morphology of Myanmar language and Bayes rule to reformulate the translation probability. Experimental results showed that the proposed system gets a BLEU-score improvement of more than 22.08% in comparison with baseline SMT system.

Informative Query Answering by using RDQL for E-Commerce

Today, the Internet can be seen as a global market place populated by a huge number of providers and consumers that exchange data from a wide range of domains. The upcoming flood of data inside the World Wide Web forces the introduction of new technologies that can reduce this information overload. Data querying is a routine practice for many on-line services. Informative query answering may improve on-line shopping service by incorporating data integration techniques. In this paper, we propose an approach for informative query answering using resource description framework query language (RDQL) and resource description framework (RDF). The RDF is used to represent all instance information. RDQL queries RDF documents using SQL-alike syntax. And the mediator/wrapper approach is used for integrating heterogeneous data

DLP Based Proxy Blind Signature Scheme with Low-Computation

Proxy blind signature which combines the properties of both proxy signature and blind signature. In a proxy blind signature scheme, the proxy signer is allowed to generate a blind signature on behalf of the original signer. The proxy blind signature scheme is useful in several applications such as e-voting, e-payment and mobile agent environments. In this paper, we present an efficient proxy blind signature scheme based on discrete logarithm problem, which satisfies the secure properties of both the blind signature scheme and the proxy signature scheme with low computation. Analysis shows that our scheme is secure and efficient.

Part of speech Tagging for Myanmar using Hidden Markov Model

Part-Of-Speech (POS) Tagging is the process of assigning the words with their categories that best suits the definition of the word as well as the context of the sentence in which it is used. In this paper, we describe a machine learning algorithm for Myanmar Tagging using a corpus-based approach. In order to tag Myanmar language, we need to take part word segmentation, part of speech tagging using HMM and several Tag-sets. Thus, this paper deals with a combination of supervised and un-supervised learning which use pre-tagged and untagged corpus respectively. To assign to each word with the correct tag, we describe Supervised POS Tagging by using the class labels in terms of predictor features on manually tagged corpus and also describe Unsupervised POS Tagging for automatically training without using a manually tagged corpus. By experiments, the best configuration is investigated on different amount of training data and the accuracy is 97.56%.