Building Common Sense Database By Using Social Media Content
Deniz Altınal, Mehmet Fatih Amasyalı · 2021 Innovations in Intelligent Systems and Applications Conference (ASYU) · 2021
In this study, it is aimed to create a dataset of Turkish concepts that contain various relations with each other by using social media contents. The problem is treated as an entity name recognition problem. Concepts and types of relationships that are related to each other are marked in the social media content used for training. The data were modeled using LSTM and BERT based methods. In this way, a large amount of unlabeled content was given to the models, and new concepts that were not in the training set and the relationships between them were found.