Osservazioni sulla Filosofia della Linguistica Computazionale. Chiarificazione dei presupposti teorici del Natural Language Processing

The article presents an overview of the literature concerning the relationship between recent Language Models and the concept of meaning. Technical advancements have prompted extensive reflections on the implications of these findings for language and semantics studies. These implications are currently fueling a lively debate among scholars across various disciplines, who are engaging in speculative discussions regarding the nature of meaning and its representation. From a philosophical perspective, the theories of meaning that emerged from such reflections often replicate several misconceptions about the nature of language outlined by Ludwig Wittgenstein in his works. The literature exhibits many of the positions criticised by the Austrian philosopher: a psychological understanding of words comprehension, a logicist interpretation of language functioning, a referentialist view of the meaning of linguistic signs, and so forth. This article endeavors to clarify these misunderstandings by drawing upon classic insights from Wittgenstein's work, in order to avoid the theoretical impasses encountered by scholars when analysing Language Models (LM). The benefits of this approach are twofold. Firstly, the phenomenon of meaning is placed in its natural context, that is language, while avoiding interference from unrelated disciplinary confusions and it becomes possible to describe the relationship between meaning and the vector representations of models.