Nurul Amelina Nasharuddin

A review on the cross-lingual information retrieval

Information retrieval involves finding some required information in a collection of information or in database. The collection not necessarily be in one language only as information does not limited to language. The simplest way to search for the information is to look at every item in the collection and when the need to translate the languages being used arises, this is where the techniques and methods that were being developed for the cross-lingual retrieval system will take place. This article reviews some recent researches focusing on topics in cross-lingual information retrieval and their role in current research directions in the wide area of information retrieval.

English and Malay cross-lingual sentiment lexicon acquisition and analysis

Sentiment analysis finds opinions, sentiments or emotions in user-generated contents. Most efforts are focusing on the English language, for which a large amount of sources and tools for sentiment analysis are available. The objective of this paper is to introduce a cross-lingual sentiment lexicon acquisition method for the Malay and English languages and further being test on a set of news test collections. Several part of speech tags are being experimented using the Word Score Summation technique in order to classify the sentiment of the news articles. This method records up to 50% as experimental accuracy result and works better for verbs and negations in both the English and Malay news articles.

Testing the accuracy of text deconstruction using PTree tool.

Our research project is currently to develop an Automatic Concept Relation Extraction (ACRE) System which automatically extracts concepts and their relationships across texts in all domains of knowledge. Concept Relational Tree (CRT) is one of the text analyzer applications used in the ACRE System to automatically extract concepts and their relationships in a document. To check on the correctness of the extraction of concepts and their relationships, the PTree is designed to reconstruct the text by reverse input. In this paper we present the PTree tool to test the accuracy of the automatic tagging and tree structure created by CRT from texts. The PTree tool is implemented from Java Universal Network/ Graph Framework (JUNG) libraries. This tool provides a few functions to allow for flexibility in drawing parse trees for concept relationships. Due to its flexibility and dynamic features, PTree can be further extended for use in the deconstruction of highly complex texts.