부산대학교 도서관

The Web has changed significantly in recent years, and websites now carry increasing information which has given rise to websites that can be difficult to navigate because of complex UI. In this work, an approach is presented which will make web navigation easier by incorporating the concepts of data scraping and intent classifiers with a natural language processing model. Our proposed concept is to create a dynamic NLUI (natural language user interface) that makes web navigation easier based on queries that can be text or voice-based. The proposed approach will help the user to obtain answers to their queries which will make web navigation easier. This is done by classifying user queries as intents and the required data is scraped from the website and when integrated with NLP models, helps achieve flexibility in understanding the context to a greater extent. The intent classification model used achieves an accuracy of 99.1 percent on our self-curated dataset. Selenium web drivers and BeautifulSoup have been used for web scraping. The integration of a BERT q/a module was also done, there are endless models and a lot of potential to scale the work and make it more flexible