Using Deep Learning for Extracting User-Generated Knowledge from Web Communities

The interest in user-generated content (UGC) is steadily growing as it provides a valuable data source for companies to extract information that can be used for competitive advantages. However, mining UGC and extracting knowledge is a costly and labour-intense endeavour. Against the backdrop, the steep advancements in Deep Learning (DL) during the last years offer the potential to counteract this. However, DL is still in its infancy in the realm of UGC. Thus, we aim at contributing to the field of knowledge extraction of UGC by comparing traditional machine learning (ML) approaches with state-of-the-art DL models (e.g., BERT) for mining user-generated instructions. We follow the knowledge discovery process to construct a novel corpus of user-generated instructions and develop a best-practice approach to mine knowledge from UGC. Thereby, we intend to foster a better understanding of knowledge extraction systems for UGC and provide a valuable solution to extend existing information systems.