Info per:
Futuri studenti
Studenti Cattolica
International Students
Docenti
Alumni
Enti, aziende e professioni
strumenti-icon
SEI UNO STUDENTE ISCRITTO?
SEI UN DOCENTE O UN MEMBRO DELLO STAFF
EN
Conferenza | 05 dicembre 2024

Natural Language Processing: From RNNs to Transformers

Brescia

Interviene

Alessandro INCREMONA
Università Cattolica del Sacro Cuore

Abstract

Natural Language Processing (NLP) is a fast-evolving field that enables machines to understand, interpret, and generate human language. This seminar explores the trajectory of NLP, tracing its development from foundational techniques to sophisticated, state-of-the-art architectures that power today’s most advanced applications. The journey begins with an introduction to core NLP tasks, such as text classification, sentiment analysis, and entity recognition, establishing the essential groundwork for language processing.
In its early stages, NLP relied on classical neural network architectures like Recurrent Neural Networks (RNNs), which were instrumental in modeling sequential data and capturing patterns in language. Despite their early success, RNNs encountered limitations, particularly with handling long sequences. This led to the development of word embeddings, techniques that represent words as dense vectors and capture semantic relationships within the language. With the advent of contextualized embeddings, models were further enhanced, enabling them to better capture subtle differences in meaning based on context.
The introduction of the encoder-decoder architecture and attention mechanisms marked a turning point in NLP, allowing models to focus selectively on relevant parts of the input and improving translation and sequence-to-sequence tasks. Building on these innovations, the Transformer architecture emerged, which abandoned recurrence in favor of self-attention, drastically improving the efficiency and effectiveness of language models.
The seminar culminates with an exploration of recent breakthroughs in NLP, such as BERT and GPT, which represent a new era in language models. These models are pre-trained on massive datasets and leverage deep contextualization, enabling them to perform complex tasks such as question answering, text generation, and more, with unprecedented accuracy and flexibility.
Through this comprehensive overview, the seminar highlights NLP’s evolution from traditional approaches to advanced architectures, shedding light on the techniques that have reshaped language technology and opened new possibilities for real-world applications.

Nome file
Locandina Natural Language Processing.pdf
Dimensione
193 KB
Formato
application/pdf
Locandina
scroll-top-icon