Shaidah Jusoh

Ambiguity in text mining

Text Mining tasks include text categorization, text clustering, concept/entity extraction, document summarization, and entity relation modeling. In this paper, the focus is given to concept/entity extraction only. The major challenging issue in extracting concept/entity from texts is natural language words are always ambiguous. Up to now, not much research in text mining especially in concept/entity extraction has focused on the ambiguity problem. This paper addresses ambiguity issues in natural language texts, and presents a new technique for resolving ambiguity problem in extracting concept/entity from texts. The technique is developed by applying possibility theory, fuzzy set, and knowledge about the context to lexical semantics.

Integrating multiple Web-based geographic information systems

Most Web-based geographic information systems (GISs) are stand-alone. Our technique aims to integrate a collection of Web GISs into a system, providing more transparent and efficient access. Users can access the data stored at any of the participating sites as though all the data came from one site. We based our technique on Common Object Request Broker Architecture (CORBA) and Java.

Distributed and Cooperative Hierarchical Intrusion Detection on MANETs

The wireless links between the nodes together with the dynamicnetwork nature of ad hoc network, increases the challenges of design and implement intrusion detection to detect the attacks. Traditional intrusion detection techniques have had trouble dealing with dynamic environments. In particular, issues such as collects real time attack related audit data and cooperative global detection. Therefore, we are motivated to design a new intrusion detection architecture which involves new detection technique to efficiently detect the abnormalities in the ad hoc networks. In this paper we present the architecture and operation of an intrusion detection technique in Mobile Ad hoc NETwork (MANET). The proposed model has distributed and cooperative architecture. The proposed intrusion detection technique combines the flexibility of anomaly detection with the accuracy of a signature-based detection method. In particular, we exploit machine learning techniques in order to achieve efficient and effective intrusion detection. A series of simulation and experimental results demonstrate that the proposed intrusion detection can effectively detect anomalies with low false positive rate, high detection rate and achieve higher detection accuracy.

Applying fuzzy sets for opinion mining

Opinions are always expressed in comments or reviews. An automated opinion mining system has been seen as one of the desirable intelligence business tools. The system can extract public opinion about a certain topic, product or service which is embedded in unstructured texts. Extracting opinions from reviews and comments requires a system to deal with natural language texts. The current approach in opinion mining is classifying sentiment words into two polar; positive and negative. Unfortunately, this is not enough. Words such as “excellent” and “good” are both classified into positive, however, the positive degree of both words are not the same. This paper introduces the use of a fuzzy lexicon and fuzzy sets in deciding the degree of positive and negative. Our experimental result shows that the approach is able to extract opinions and present the opinions in a more efficient way.

Automated Translation Machines: Challenges and a Proposed Solution

Automated translation (MT) tools have become an urgent need in a multilingual environment. Although there are any available tools on the market, unfortunately, a robust MT tool is still a dream. This purpose of this paper is to discuss challenging issues in MT tool developments, the state of art of he MT tools and propose a framework for a semantic-based translation. The focus of this paper is English to Arabic translation MT.

Handling imbalance visualized pattern dataset for yield prediction

The prediction of the yield outcome in a non close loop manufacturing process can be achieved by visualizing the historical data pattern generated from the inspection machine, transform the data pattern and map it into machine learning algorithm for training, in order to automatically generate a prediction model without the visual interpretation needs to be done by human. Anyhow, the nature of manufacturing process dataset for the bad yield outcome is highly skewed where the majority class of good yield extremely outnumbers the minority class of bad yield. Comparison between the undersampling, over- sampling and SMOTE + VDM sampling technique indicates that the combination of SMOTE + VDM and undersampled dataset produced a robust classifier performance capable of handling better with different batches of prediction test data sets. Furtherance, suitable distance function for SMOTE is needed to improve class recall and minimize overfitting whilst different approach on the majority class sampling is required to improve the class precision due to information loss by the undersampling.

Automated Text Summarization: Sentence Refinement Approach

Automated text summarization is a process of deriving a shorter version of a text document from an original text. The most well known and widely used technique for automated text summarization is sentence extraction technique. Using this technique, sentences are extracted based on certain features that have been decided. In this paper, a new technique called sentence refinement is introduced as an improvement of the technique. In this approach, a sentence is refined; unimportant words or phrases exist in the extracted sentences are omitted. A summarization tool has been developed based on the proposed approach. The tool was tested using English and Malay texts. Extrinsic and intrinsic measurement methods have been used in evaluating generated summaries. Results show the proposed approach is promising.

Visualizing the Yield Pattern Outcome for Automatic Data Exploration

Non close loop manufacturing process, typically in the hard disk media industries rely from its inspection machine to generate production yield temporal data that can be used for future analysis. In order for an engineer to proactively perform maintenance on its process equipment and avoiding unnecessary unplanned down time, they need to be able to predict the outcome of the yield before products arrives at the inspection machine. The future prediction of the yield outcome can be achieved by visualizing the historical data pattern generated from the inspection machine, transform the data pattern and map it into machine learning algorithm for training in order to automatically generate a prediction model without the visual interpretation needs to be done by human.

Resolving ambiguous preposition phrase for text mining applications

Text Mining is one of the computational intelligence research areas. The main goal of text mining tool is to discover knowledge which is embedded in unstructured text. The first step of text mining is to extract fact from the texts. However, to build a robust text mining tool is very complex. The first step requires the tool to process a natural language. The major challenging issue in any natural languages is the ambiguity problem. The problem may occur at lexical and phrase levels. This paper addresses ambiguity problem which occur in the preposition phrase, and presents a new technique for resolving the problem. The technique has been developed by applying possibility theory, fuzzy set, and context knowledge. The technique has been implemented and tested using a set of test cases and promising results are obtained.

Visualizing the Yield Pattern for Multi Class Classification

This research attempts to generate an automatic prediction model in a hard disk media manufacturing process. This is to be done without human visual interpretation. Our research demonstrates that it can be achieved by visualizing the historical temporal data pattern generated from the inspection machine. From there, the data pattern is transformed and mapped into machine learning algorithm for training. In this paper, we have introduced the pattern visualization technique with trinary and quinary number and compared them with our previous binary pattern visualization technique. This is to deal with multi class classification. The result implied that, the performance of the multi class classification can be improved when all class instances were made higher in quantity and balance. Quinary pattern visualization techniques performed better compared with binary and trinary patterns when the multi class instances were made balanced and were significantly at higher quantity.