D
Danilo Croce
Researcher at University of Rome Tor Vergata
Publications - 108
Citations - 1684
Danilo Croce is an academic researcher from University of Rome Tor Vergata. The author has contributed to research in topics: Sentiment analysis & Tree kernel. The author has an hindex of 21, co-authored 102 publications receiving 1437 citations. Previous affiliations of Danilo Croce include Sapienza University of Rome.
Papers
More filters
Proceedings Article
Structured Lexical Similarity via Convolution Kernels on Dependency Trees
TL;DR: This paper defines efficient and powerful kernels for measuring the similarity between dependency structures, whose surface forms of the lexical nodes are in part or completely different, and confirms the benefit of semantic smoothing for dependency kernels.
Proceedings ArticleDOI
GAN-BERT: Generative adversarial learning for robust text classification with a bunch of labeled examples
TL;DR: This paper proposes GAN-BERT that ex- tends the fine-tuning of BERT-like architectures with unlabeled data in a generative adversarial setting, and shows that the requirement for annotated examples can be drastically reduced.
Book ChapterDOI
EVALITA 2020: Overview of the 7th Evaluation Campaign of Natural Language Processing and Speech Tools for Italian.
TL;DR: The tasks proposed at EVALITA 2020 are introduced and an overview to the participants and systems whose descriptions and obtained results are reported in these Proceedings is provided.
Book ChapterDOI
Overview of the Evalita 2016 SENTIment POLarity Classification Task
TL;DR: The datasets are presented – which includes an enriched annotation scheme for dealing with the impact on polarity of a figurative use of language – the evaluation methodology is discussed, and results and participating systems are discussed.
Proceedings ArticleDOI
KeLP at SemEval-2016 Task 3: Learning Semantic Relations between Questions and Answers
TL;DR: This paper describes the KeLP system participating in the SemEval-2016 Community Question Answering (cQA) task, which outperforms all the other systems with respect to all theother challenge metrics.