Provision of a new dataset to this field to work on it.  Showing the effects of noise normalization and preprocessing on the classification accuracy.  Representation of texts in vector form in various ways to be able to work on them.  Comparison of the LSTM Network with several machine learning algorithms.

In today's conditions, it is difficult to obtain information quickly and efficiently due to the size of the data. There are various text documents on the internet and a good extraction algorithm is essential to have the most relevant information from them. Long texts can be boring sometimes. So, readers are eager to get the main idea of the text or any useful information. For this reason, the importance of automatic summarization systems is understood. Text summarization systems can be considered as abstractive summarization or extractive summarization. While abstractive systems produce a summary with new sentences, extractive systems make a selection of sentences from the text used and combine them and present them as a summary. Creating a successful summarization algorithm increases in direct proportion to the success of applying text mining techniques. Text summary systems provide a summary of the text to the user by scoring words and sentences in the main text using various methods and combining high ranked sentences as a result of the process. In this context, many scoring methods have been used. In our study, news data sets are used. The algorithm used is based on extraction and has been evaluated using a taskindependent method. After evaluation, the two highest scores taken are ROUGE-1 with 0.68 score and ROUGE-S with 0.54 score. Through all evaluation steps, Precision, Recall and F-Measure values are also specified to see the steps clearly.