Covid-19 Semantic Browser
In 2020, more than 3,000 scientific studies have been published on the SARS-CoV-2 virus and on the Covid-19 pathology. The total number of articles on the topic of coronaviruses exceeds 40,000 units. Such a volume of scientific production makes it impossible for doctors and researchers to keep up with the latest discoveries without the support of adequate digital platforms that are currently nowhere in sight.
To make up for this shortcoming, we propose an artificial intelligence system associated with a web application to perform natural language semantic querying inside the COVID-19 Open Research Dataset published by the American nonprofit AllenAI. The system leverages state-of-the-art neural language models trained on scientific publications in the biomedical domain for optimal retrieval performances. The adoption of the system aims to facilitate knowledge sharing across the scientific community and to accelerate the development of adequate drugs and vaccines to counter the ongoing pandemic.
The project is led by Gabriele Sarti in collaboration with Area Science Park and the Italian Association of Computational Linguistics. The project used to be publicly available at covidbrowser.areasciencepark.it. You can now refer to the code implementation on GitHub.
- Neural Language Models: the New Frontier of Natural Language Understanding
- Probing Linguistic Knowledge in Italian Neural Language Models across Language Varieties
- Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks
- The Literary Ordnance: When the Writer is an AI
- Histopathologic Cancer Detection with Neural Networks