Lavagna per la pace e Natural Language Processing

This study has a threefold objective: a) to verify consistency between recurrent terms and the value context; b) to discover latent topics of peace; and c) to verify the syntactic invariance of the peace construct by comparing two subgroups characterized by distinct sensitivities: subjective and objective. For the first objective, a singularity has been captured: the terms that normally have a positive valence are placed in a negative value context, and this confirms the need to develop more and more analytical methodologies capable of combining the analysis of linguistics with that of textual statistics since the simple BoW (Bag of Words) approach may not faithfully reproduce the value context in a scenario of small size and short texts. The use of latent topic modeling allows identification of the underlying topics for the second goal, and the Glove algorithm, developed for similarity detection, has been used for the third aspect. In detail, the freedom construct was highlighted in the subjective group, while for the objective group, the most characterizing similarity was identified in the lemma vivere.