Progressive Guidance Categorization Using Transformer-Based Deep Neural Network Architecture

In the era of advanced technology, the internet is one of the most indispensable necessities in our daily liveliness. Usually, people search for various things on a search engine, from daily needs to dream utilities. From the ample amount of searching styles, the prefix with "How To" is one of the exploited prominent styles. According to this search style with the mentioned prefix, people mainly find out the guidelines or instructions for solutions. However, there is no work has been conducted on guidelines or instructions classification in the past. In this research work, we identify the category of guidelines or instructions from the summary of "How To" articles. We employ some deep neural architectures, such as LSTM (Long Short-Term Memory), RNN (Recurrent Neural Network), and BERT (Bidirectional Encoder Representations from Transformers) to classify guidelines precisely and swiftly. We have conveyed this work with a novel dataset comprising 11,121 observations from the WikiHow. In this classification process, we have manifested accuracy, precision, recall, and fl-score to appraise our proposed architectures. The outcomes have brought a wit of our applied BERT architecture that achieves notably with 87% accuracy.