نوع مقاله : مقاله پژوهشی
نویسندگان
1 کارشناس ارشد کامپیوتر، دانشگاه گلستان، گرگان
2 استادیار گروه کامپیوتر، دانشگاه گلستان، گرگان
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
The purpose of this study is to verify combination of some classification and summarization techniques and to examine evaluation metrics of classification. The proposed framework implemented in seven main stages. First, 1,000 documents collected from yjc.ir website. The selection of documents is based on the appropriate content and a minimum of 100 and a maximum of 350 words. These documents divided into three categories: document title, document summary and original text of the document. Summary text and the original text grouped into 250, 500 and 1000 documents in two stages, with a 100% growth in the number of documents. The pre-processing of text performed and the stop-words deleted from the sentences. Next, the TF-ISF summarizer techniques implemented. A variety of classification algorithms such as Decision trees, Support vector machine, Bayesian and Rule implemented by the RapidMiner software, which provided 120 Excel outputs from the results of the evaluation criteria (accuracy, precision, and recall). Finally, five comparisons between the results considered. The results of this study indicate that the superiority of 1,000 documents, the ISF summarizer method versus TF, Bayesian and SVM classification versus Rule and Decision tree classifications, the original text versus summary text with highest of 96.67% of accuracy in SVM classification, 1000 documents and ISF summarizer technique.
کلیدواژهها [English]
ارسال نظر درباره این مقاله