The Dark Side of the Language: Pre-trained Transformers in the DarkNet

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Ranaldi, Leonardo, Nourbakhsh, Aria, Patrizi, Arianna, Ruzzetti, Elena Sofia, Onorati, Dario, Fallucchi, Francesca, Zanzotto, Fabio Massimo
Format: Preprint
Publié: 2022
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866916234658840576
author Ranaldi, Leonardo
Nourbakhsh, Aria
Patrizi, Arianna
Ruzzetti, Elena Sofia
Onorati, Dario
Fallucchi, Francesca
Zanzotto, Fabio Massimo
author_facet Ranaldi, Leonardo
Nourbakhsh, Aria
Patrizi, Arianna
Ruzzetti, Elena Sofia
Onorati, Dario
Fallucchi, Francesca
Zanzotto, Fabio Massimo
contents Pre-trained Transformers are challenging human performances in many NLP tasks. The massive datasets used for pre-training seem to be the key to their success on existing tasks. In this paper, we explore how a range of pre-trained Natural Language Understanding models perform on definitely unseen sentences provided by classification tasks over a DarkNet corpus. Surprisingly, results show that syntactic and lexical neural networks perform on par with pre-trained Transformers even after fine-tuning. Only after what we call extreme domain adaptation, that is, retraining with the masked language model task on all the novel corpus, pre-trained Transformers reach their standard high results. This suggests that huge pre-training corpora may give Transformers unexpected help since they are exposed to many of the possible sentences.
format Preprint
id arxiv_https___arxiv_org_abs_2201_05613
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle The Dark Side of the Language: Pre-trained Transformers in the DarkNet
Ranaldi, Leonardo
Nourbakhsh, Aria
Patrizi, Arianna
Ruzzetti, Elena Sofia
Onorati, Dario
Fallucchi, Francesca
Zanzotto, Fabio Massimo
Computation and Language
Machine Learning
Pre-trained Transformers are challenging human performances in many NLP tasks. The massive datasets used for pre-training seem to be the key to their success on existing tasks. In this paper, we explore how a range of pre-trained Natural Language Understanding models perform on definitely unseen sentences provided by classification tasks over a DarkNet corpus. Surprisingly, results show that syntactic and lexical neural networks perform on par with pre-trained Transformers even after fine-tuning. Only after what we call extreme domain adaptation, that is, retraining with the masked language model task on all the novel corpus, pre-trained Transformers reach their standard high results. This suggests that huge pre-training corpora may give Transformers unexpected help since they are exposed to many of the possible sentences.
title The Dark Side of the Language: Pre-trained Transformers in the DarkNet
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2201.05613