Quotations identification from Indonesian online news using rule-based method

摘要

Interviewee's statement is one of the source of information in news. The statement may affect the mind of the reader and may also indicate sentiment expression related to the subject of interest. Expressions of the interviewee are often written using quotation, either direct or indirect. This paper will provide a discussion on how to identify a quotation from an Indonesian online news text using rule-based method. Each pattern structure of writing quotations commonly found in news will be made into a class rule that represents the structure. Features used in making of class rule are the existence of an entity and reporting verb in the sentence. The result indicates that the entire class rules that have been made are able to extract sentence statement pretty well. The system is able to detect sentence statement with an accuracy of 88.618%, precision of 99.013% and recall 79.936%.