Development of recommendation system for pupil’s Informal education based NLP and LSTM network

Abstract

Data from international pupil achievement surveys show that the general learning outcomes of Lithuanian pupils in the international context remain quite average (EBPO PISA, 2018), while the national achievements of pupils are also not high. It is important to look for a variety of tools that can help students achieve higher learning outcomes. Non-formal education is one of the directions that help to solve this problem. This research is looking for recommendation systems based on machine learning models that can recommend to the pupil the non-formal education field to improve learning outcomes. A generator of recommendations for informal education services has been developed, forming personal recommendations for users of formal education based on the newly developed algorithms of machine learning. The recommendation generator consists of parts such as a data preparation aggregator that collects and transforms data. One of the main parts of the model is the classification of text-based feedback messages from teachers into established categories using the NLP (Natural language processing) technique. A multi-layered LSTM network is trained to prepare the data, with the help of which a recommendation for the non-formal learning direction is subsequently provided. The developed non-formal science recommendation generator is available as a tool that can help students achieve higher learning outcomes