Abstract:

Natural Language Processing (NLP) is a computerised approach to text analysis based on a set of theories and technologies; this Artificial Intelligence (AI) algorithm should be able to analyse, represent and understand human language. Thanks to the contribution of more and more advanced AI techniques, such as Machine Learning and Deep Learning, it is possible to find various fields of application for NLP, even considering the complexity of a language like Italian, characterised by idioms, slang expressions, metaphors and numerous dialects. The dialogue between man and machine inevitably involves various aspects of communication and discourse as a whole (semantics, pragmatics, syntax, etc.). Tasks that automate these areas involve simple tasks such as language recognition, sentence decomposition into elementary units, semantic analysis and sentiment analysis.
Given this, we intend to reflect on a hypothetical impact of AI on learning environments, to highlight desirable didactic implications; pupils' feedback, regarding the services offered by educational infrastructures, could be analysed using NLP techniques, identifying specific areas of improvement in good practice. Specifically, the study aims at proposing a pedagogical model that supports a renewal of didactics, through the identification of NLP methodologies and applications, to explore the semantic meaning of feedback and emerging sentiments and promote, at the same time, the process of cultural and social accessibility, with a view to quality, equitable and inclusive education that should leave no one behind (Goal 4 - UN Agenda 2030).
The proposed didactic framework exploits the potential of NLP to make the text usable, replacing more elaborate constructs with simpler words and adapting the text about the specificity of the pupil; in this way, a more streamlined and intuitive version of the text can be generated, which will be differentiated in
the case of pupils with reading and comprehension difficulties. In particular, natural language processing algorithms are used to analyse the original text and identify passages that might be difficult for students. These passages are then simplified and made more accessible, without compromising the conceptual integrity. Furthermore, the text is dynamically adapted to the specific needs of the student, creating a customised version that facilitates comprehension. The
study aims to analyse the effectiveness of such processes by providing students with the original version of a text and its simplified version, and analysing through sentiment analysis techniques the students' responses to improve the scalability and dynamism of the process. The use of such a framework in terms
of results appears satisfactory and encourages the continuation of the analysis with further expedients such as the use of text classification techniques as a preparatory process to the framework itself.

Keywords:

Artificial Intelligence, Natural Language Processing, Inclusion, Learning environments.