Turkish Question answering sentiment dataset (SCD)
SCD Dataset: This new dataset has been specifically created for the development and education of children with down syndrome. The dataset, containing a total of 13,500 Turkish question-answer pairs, has "positive" and "negative" emotion labels. In the context of human-robot interaction, accurately identifying and addressing positive and negative emotions has a significant impact on user experience and satisfaction. Neutral questions and answers provide less information in terms of sentiment analysis and are less relevant to the purpose of this study. Therefore, focusing on positive and negative emotions in our research was preferred in order to achieve more specific and effective results. In this context, situations of neutral questions and answers that approach positive or negative meanings have been appropriately manually corrected. The dataset has been carefully and originally prepared, based on various topics that can be covered in conversations between robots and children. In this dataset, the question class can include any type of sentence, whether it is a single word or meant to maintain the continuity of the conversation. Similarly, the answer class is designed in the same manner. The entire set of question-answer pairs, created with 200 volunteer students from the Computer Engineering Department of Iskenderun Technical University, has been labeled with emotion tags and verified for accuracy by experts in the field.