Performance Evaluation of SVM-Based Information Extraction using τ Margin Values

The rapid growth of Internet causes the abundance of textual information. It is necessary to have smart tools and methods than can access text content as needed. One of the success methods is Support Vector Machine (SVM). This paper will discuss how the performance of the SVM-GATE algorithm on extracting information from Indonesian language corpus in response toτ margin variation. Experimental results show that there is optimumτ margin for both Indonesian corpus of Vegetable Market and Seminar Announcement Corpus. The best Performance of SVM-GATE obtained at the τ Margin of 0.5 and the Window Size of 4x4.