Chuleerat Jaruskulchai

Generic text summarization using local and global properties of sentences

With the proliferation of text data on the World-Wide Web, the development of methods for automatically summarizing these data becomes more important. Here, we propose a practical approach for extracting the most relevant sentences from the original document to form a summary. The idea of our approach is to exploit both the local and global properties of sentences. The local property can be considered as clusters of significant words within each sentence, while the global property can be though of as relations of all sentences in the document. These two properties are combined to get a single measure reflecting the informativeness of sentences. Experimental results show that our approach compares favorably to a commercial text summarizer.

Band Selection for Dimension Reduction in Hyper Spectral Image Using Integrated Information Gain and Principal Components Analysis Technique

Nowadays, hyper spectral image software be- comes widely used. Although hyper spectral images provide abundant information about bands, their high dimensionality also substantially increases the computational burden. An important task in hyper spectral data processing is to reduce the redundancy of the spectral and spatial information without losing any valuable details. In this paper, we present band selection technical using principal components analysis (PCA) and information gain (IG) for hyper spectral image such as small multi-mission satellite (SMMS). Band selection method in our research not only serves as the first step of hyper spec- tral data processing that leads to a significant reduction of computational complexity, but also a invaluable research tool to identify optimal spectral for different satellite applications. In this paper, an integrated PCA and IG method is proposed for hyper spectral band selection. Based on tests in a SMMS hyper spectral image, this new method achieves good result in terms of robust clustering.

A parallel learning algorithm for text classification

Text classification is the process of classifying documents into predefined categories based on their content. Existing supervised learning algorithms to automatically classify text need sufficient labeled documents to learn accurately. Applying the Expectation-Maximization (EM) algorithm to this problem is an alternative approach that utilizes a large pool of unlabeled documents to augment the available labeled documents. Unfortunately, the time needed to learn with these large unlabeled documents is too high. This paper introduces a novel parallel learning algorithm for text classification task. The parallel algorithm is based on the combination of the EM algorithm and the naive Bayes classifier. Our goal is to improve the computational time in learning and classifying process. We studied the performance of our parallel algorithm on a large Linux PC cluster called PIRUN Cluster. We report both timing and accuracy results. These results indicate that the proposed parallel algorithm is capable of handling large document collections.

A Practical Text Summarizer by Paragraph Extraction for Thai

In this paper, we propose a practical approach for extracting the most relevant paragraphs from the original document to form a summary for Thai text. The idea of our approach is to exploit both the local and global properties of paragraphs. The local property can be considered as clusters of significant words within each paragraph, while the global property can be though of as relations of all paragraphs in a document. These two properties are combined for ranking and extracting summaries. Experimental results on real-world data sets are encouraging.

Possibilistic Exponential Fuzzy Clustering

Generally, abnormal points (noise and outliers) cause cluster analysis to produce low accuracy especially in fuzzy clustering. These data not only stay in clusters but also deviate the centroids from their true positions. Traditional fuzzy clustering like Fuzzy C-Means (FCM) always assigns data to all clusters which is not reasonable in some circumstances. By reformulating objective function in exponential equation, the algorithm aggressively selects data into the clusters. However noisy data and outliers cannot be properly handled by clustering process therefore they are forced to be included in a cluster because of a general probabilistic constraint that the sum of the membership degrees across all clusters is one. In order to improve this weakness, possibilistic approach relaxes this condition to improve membership assignment. Nevertheless, possibilistic clustering algorithms generally suffer from coincident clusters because their membership equations ignore the distance to other clusters. Although there are some possibilistic clustering approaches that do not generate coincident clusters, most of them require the right combination of multiple parameters for the algorithms to work. In this paper, we theoretically study Possibilistic Exponential Fuzzy Clustering (PXFCM) that integrates possibilistic approach with exponential fuzzy clustering. PXFCM has only one parameter and not only partitions the data but also filters noisy data or detects them as outliers. The comprehensive experiments show that PXFCM produces high accuracy in both clustering results and outlier detection without generating coincident problems.

A Simple Approach to Optimize Web Services' Performance

The standard of data transfer and exchange between organizations through Web services has become exceedingly popular, especially in electronic commerce. Web services are somewhat loosely defined, but may be characterized in general as using existing web technologies and standards to build the distributed computing environments. Data transferred through web services is in the form of XML. Even though Web services can be beneficial and so it influences the effectiveness of the system as a whole. In this paper presents algorithm in optimizing the effectiveness of data transferring through the network using file transfer protocol. The mechanism has been reduces the data exchange time up to 70%. In addition, it can also send data through network with efficiency without losing standards of data transfer and avoid traffic from HTTP protocol.

Determination of the Initialization Number of Clusters in K-means Clustering Application Using Co-Occurrence Statistics Techniques for Multispectral Satellite Imagery

 Abstract—Nowadays, clustering is a popular tool for explo- ratory data analysis, such as K-means and Fuzzy C-mean. Automatic determination of the initialization number of clus- ters in K-means clustering application is often needed in ad- vance as an input parameter to the algorithm. In this paper, a method has been developed to determine the initialization number of clusters in satellite image clustering application using a data mining algorithm based on the co-occurrence matrix technique. The proposed method was tested using data from unknown number of clusters with multispectral satellite image in Thailand. The results from the tests confirm the effectiveness of the proposed method in finding the initializa- tion number of clusters and compared with isodata algorithm. Clustering is a popular tool for data mining and explora- tory data analysis. One of the major problems in cluster analysis is the determination of the number of clusters in unlabeled data, which is a basic input for most clustering algorithms. In this paper, we propose a new easy method for automatically estimating the number of clusters in unlabeled data set. Pixel clustering technique in a color image is a process of unsupervised classification of hundreds thou- sands or millions pixels on the basis of their colors. One of the most popular and fastest clustering techniques is the k- means technique. The results of the k-means technique depend on different factors such as a method of determina- tion of initial cluster centers as shown in Fig. 1. Such sensi- tivity to initialization is an important disadvantage of the k- means technique. In this paper, a method has been devel- oped to determine the initialization number of clusters in satellite image clustering application using a data mining algorithm based on the co-occurrence matrix technique. Therefore, automatic determination of the initialization number of clusters can greatly help with the unsupervised classification of satellite Image.

XML Retrieval More Efficient Using ADXPI Indexing Scheme

XML Information Retrieval is approach to identify the appropriate answer granularity and controlling to elements overlap. Recently, the demand for integrating Full Text Search and relational search has increased dramatically. The RDBMS implementation is generally much worse in the performance than the IR engine implementation. Especially, when a query is processed in the RDBMS, the number of join operation increases in proportion to the number of relationships in the query. In order to solve these problems, we propose in this paper a novel approach to extend the inverted index for support query processing, namely Absolute Document XPath Indexing that allows supporting and reducing the length of time on Score Sharing scheme. In terms of processing time, our system required an average of one second per topic on INEX-IEEE and an average of ten seconds per topic on INEX-Wiki better than GPX system.

A simple approach to optimize XML Retrieval

In this paper, we report experimental results of our approach using BM25E model for retrieval large-scale XML collection, to improve the effectiveness of XML Retrieval. This model is commonly used in the information retrieval community. We propose new algorithm using Score Sharing that allow to assign parent score by sharing score from leaf node to their parents by a Top-Down Scheme approach. In order to improve efficiency on response time, The Score Sharing algorithm processing time on 10,000 leaf nodes is around 0.135 ms. per topic after getting the result list from Zettair. The Zettair is able to process on average time per topic using less than 1 second then the processing time is up to 1 second per topic and our experiment show that the BM25E with Score Sharing improve iP[0.10] by 24.40% and MAiP by 31.89% over the original BM25E. In addition, our algorithm able to handle both elements level and document level by only setting parameter.

Membership enhancement with exponential fuzzy clustering for collaborative filtering

In Recommendation System, Collaborative Filtering by Clustering is a technique to predict interesting items from users with similar preferences. However, misleading prediction could be taken place by items with very rare ratings. These missing data could be considered as noise and influence the clusters centroid by shifting its position. To overcome this issue, we proposed a new novel fuzzy algorithm that formulated objective function with Exponential equation (XFCM) in order to enhance ability to assign degree of membership. XFCM embeds noise filtering and produces membership for noisy data differently to other Fuzzy Clustering. Thus the centroid is robust in the noisy environment. The experiments on Collaborative Filtering dataset show that centroid produced by XFCM is robust by the improvement of prediction accuracy 6.12% over Fuzzy C-Mean (FCM) and 9.14% over Entropy based Fuzzy C-Mean (FCME).