News

• 01-01-2021 New multilingual evaluation suite for Word Sense Disambiguation! Meet XL-WSD, accepted at AAAI2021.
• 01-11-2020 New multilingual evaluation suite for the Word in Context task. Meet XL-WiC, accepted at EMNLP2020.
• 01-11-2020 New English and Multilingual embeddings for word senses, i.e., all WordNet nodes and their corresponding in other languages! All thanks to ARES, check it out at EMNLP2020!
• 12-07-2020: A survey on The Knowledge Acquisition Bottleneck Problem in Multilingual Word Sense Disambiguation is now published in the IJCAI2020 proceedings.
• 12-07-2020: MuLaN, a multilingual approach to transfer semantic annotations from English to other languages (tested in DE ES FR and IT) has been published at IJCAI2020.
• 05-05-2020: CluBERT accepted at ACL2020. CluBERT is a cluster-based approach for automatically inducing the distribution of word senses from a corpus of raw texts. We tested on 5 different languages, i.e., English, French, German, Italian and Spanish and release the distributions of lexemes in the test sets at https://github.com/SapienzaNLP/clubert. More data will be soon released!
• 12-02-2020: Poster and slides of SensEmBERT AAAI2020 paper.
• 12-02-2020: Poster of CSI AAAI2020 paper.
• 10-01-2020: I am part of the organizing committee of EurNLP 2020!
• 22-11-2019: AAAI 2020 papers are out!!!
  • SensEmBERT:Context-Enhanced Sense Embeddings for Multilingual Word Sense Disambiguation, joint work with Bianca Scarlini and Roberto Navigli. We present a knowledge-based approach for producing sense embeddings in multiple languages that lay in the same semantic space of BERT contextualised embeddings. Our embeddings achieve SOTA results on all WSD datasets for English and 4 other languages. Stay tuned for website and data!
  • CSI: A Coarse Sense Inventory for 85% Word Sense Disambiguation, joint work with Caterina Lacerra, Michele Bevilacqua and Roberto Navigli. We present a new organization of concepts based on a large-scale mapping of WordNet synsets to domain-based semantic labels. Our new coarse-grained inventory of semantic labels proved to be more descriptive for humans than other competitors and aided a BERT-based WSD model to attain more than 85% accuracy.
    Our new coarse-grained inventory along with the mapping to WordNet synsets will be available soon!
• 16-11-2019: Two papers accepted at AAAI 2020 !!**: SensEmBERT with Bianca Scarlini and Roberto Navigli and CSI with Caterina Lacerra, Michele Bevilacqua and Roberto Navigli. Data and preprints will be available soon. In the meanwhile, check out my tweet for a breef description of the two papers!
• 18-10-2019: Invited talk at Huawei in Helsinki on the application of knowledge bases to down-stream NLP applications.
• 17-10-2019: Invited talk at University of Helsinki on my past, present and future research.
• 12-10-2019: Attended the first EurNLP conference together with Bianca Scarlini where we presented OneSeC!
• 01-08-2019: New paper out and published at ACL 2019 with Bianca Scarlini and Roberto Navigli. OneSeC: Automatically generated data annotated with word meanings in multiple languages. Check it out and download the data from our website!
• SemEval 2018 shared task on hypernym discovery.
• Here there are a bunch of Java utilities I’m developing during my phd in NLP in order to make my life simpler.
• Enjoy WiBi web interface, an interface to navigate our bi-taxonomy of Wikipedia pages and Categories in mutliple languages.
• Heap C implementation on github! Enjoy and please contact me for any problem or advice.