DEVELOPING TOPICAL CLUSTERING OF LIBYAN’S TWEETS USING CARROT2 FRAMEWORK

Abstract

Today, Twitter in addition has become a way of life also has become one of the sources of data in a very large number that can be accessed by anyone and anytime as long as it is connected to the internet. Data twitter is now widely used for various purposes, e.g. for the purposes of the survey. As we know that today many organizations, both government and non-government have been utilizing Twitter as a data source. In this study, developed an application to analyze data or tweets twitter committed by the Libyan people, where most of the tweets are written in letters and Arabic. The results of the analysis are expected to be topics of conversation among the Libyan people on Twitter.

Development is done by utilizing Carrot² Framework, where the stages of analysis and design are done by using the object-oriented approach utilizing UML. In addition, the application also utilizes several open libraries, such as Apache Lucene is useful for the process of indexing and retrieval and also Twitter4J tweets, a useful library API to access tweets from Twitter.

The resulting application is then tested with multiple queries, and the results of these tests can then be calculated degree of accuracy in the system generate topics of conversation.  With 10 queries, system generates 197 topics, which 185 topics are relevant while the 12 topics considered irrelevant. So the accuracy of the system can be calculated and the results obtained 93.91. It can be concluded that the accuracy of the system in generating relevant topics is 93.91%.