Exploring Cross-Lingual Named Entity Recognition: A Study of the ConNER Model for the Italian Language

Abstract

This dissertation provides a comprehensive overview of Named Entity Recognition (NER) in Natural Language Processing (NLP), encompassing its historical context, key principles, and cutting-edge techniques. It focuses on cross-lingual NER models, exploring how they leverage shared knowledge among languages to enhance performance.

We investigate the advantages and limitations of cross-lingual NER, considering reduced data annotation and improved generalization and their challenges, including language variations and resource availability.

A pivotal aspect is the analysis of ConNER, a state-of-the-art cross-lingual NER model, with a focus on its performance in the Italian language. Our empirical study employs a modified MultiNERD dataset covering English, German, French and Spanish, shedding light on ConNER's adaptability to other languages.

Ultimately, this research aims to enrich NER methodology, offering insights into the potential of cross-lingual approaches for improving NER systems.