Emotion detection and analysis from the data obtained from the speeches in Turkish audio recordings

Emotions are physical changes in a person's mood resulting from his interaction with internal and environmental influences. Emotions or emotions can be conveyed by voices and speeches besides gestures and gestures. Emotion detection in sound has become important with the developments in data science. Especially at times and places where body language is not effective, for example, in data such as voice recordings, the emotion detection and analysis made from speaking has risen to the top in the subjects of study about this field with the developing technology. Emotion detection is expressed in two ways. The first of these is anger, happiness, sadness, etc., which express more human emotions. are emotions. The other emotion expression form is the positive term that expresses positive situations such as joy and happiness, the negative term that expresses negative situations such as anger and unhappiness, and the neutral term used for situations that do not cause any reaction in the person. Emotional transfers expressed with the terms positive, negative and neutral are considered more suitable for machine language. Besides machine learning models to teach human emotions to machine language, artificial intelligence algorithms are also used. In this way, in addition to emotion detection, emotion detection is analyzed with machine learning models. In this study, it is aimed to create a Turkish emotion database by performing emotion detection and analysis of a data set consisting of Turkish voice recordings, and to be a pioneer in other Turkish studies with the content and method of the study. The dataset used in the study was obtained from Common Voice open-source audio datasets. Emotion detection was not carried out directly through the voice recordings in the data set, the voice recordings were converted into text, and emotion detection was carried out through these texts. This was one of the things that made the study different. Two different variations of the BERT model, and the TextBlob library were used for emotion detection. Another important point of the study was the analysis made with machine learning models after the emotion detection on the data set. After the literature research, these analyzes were carried out with the five most used machine learning models in studies. Analyzes were performed in both Python programming language and RapidMiner data analysis platform. The fact that the feeling of emotion in the study has been issues that make the fact that the analysis is organized in two different ways and in their own.