Bridging Theoretical Linguistics and Automated Language Processing: Emerging Synergies and Advances

In recent years, Natural Language Processing (NLP) has gone through a radical breakthrough due to the success of Large Language Models (LLMs). These models have demonstrated that an unsupervised inductive approach, based on processing large amounts of raw data, can achieve extraordinary accuracy rates in many NLP-related tasks. This progress seems to have reduced the role of theoretical linguistics in computational approaches to linguistic analysis. For instance, metalinguistic annotation, a central element in supervised learning, which traditionally relies on established linguistic theories, has lost its core place in developing trained models for different NLP tasks. In this context, a closer dialogue between theoretical and computational linguistics becomes essential to integrate theoretical and practical perspectives, bridging the gap between these two approaches to language study. This is particularly important for both the practical implementation of linguistic theories in computational models and the theoretical interpretation of results produced by NLP systems. This special issue of the Italian Journal of Computational Linguistics aims to bridge this gap by exploring the emerging synergies between theoretical linguistics and computational linguistics and the implications these have for the future of the two approaches.