Sentiment analysis using artificial intelligence methods for Turhish texts

Sentiment analysis is the determination of people's emotional state using various methods. When performed within a text, it can be expressed as the determination of the opinions shared within a text using various methods. Many problems related to grammar rules are encountered during the sentiment analysis of texts written in daily spoken language. Sentiment analysis can be done at text-based or target-based level. With text-based sentiment analysis; the general sentiment of the text can be determined. However, in this case, some sentiments may be lost. In order to prevent the loss of sentiment, with target-based sentiment analysis; different categories can be determined within the text and the sentiment status can be determined separately for each category. In this thesis study, both text-based sentiment analysis and target-based sentiment analysis were performed according to positive, negative and neutral sentiment classes using Turkish text data. For text-based sentiment analysis, the sentiment status of each comment in the data set was determined separately using machine learning algorithms. For target-based sentiment analysis, the target categories in each comment were first determined using the proposed method. Then, target-based sentiment analysis was performed at 2 different levels: sentiment analysis of the text according to the determined target categories and sentiment analysis of the data set according to the target categories. All these applications were performed on the Weka platform using different machine learning algorithms, training and testing dataset method and k-layer cross validation method. As a result of the study, the results were examined according to success criteria such as accuracy rate, area under receiver operating characteristic, F-Measure, time taken to train the model.