Deep learning for next-generation biomedical text mining

Description of the granted funding

This project will combine the entire publicly available biomedical domain scientific literature of over 10 billion words, analyses from three literature-scale text mining resources, and the latest deep neural network-based machine learning methods for language processing to create the next generation of biomedical text mining methods. We will merge three independently created analyses of the literature and find texts that are annotated identically in all three, i.e. where all agree on for example what genes, proteins are mentioned, which of them are stated to be related, and how. In this way, we will create a high-quality dataset of hundreds of millions of high-quality analyses, enough to train very deep and data-hungry neural network models to perform such annotation automatically with high accuracy. These models will then be used to re-analyze the literature, producing a molecular network model capturing more domain knowledge from the scientific literature than previously possible.
Show more

Starting year

2020

End year

2024

Granted funding

Sampo Pyysalo Orcid -palvelun logo
532 345 €

Funder

Research Council of Finland

Funding instrument

Academy projects

Other information

Funding decision number

332844

Fields of science

Computer and information sciences

Research fields

Tietojenkäsittelytieteet

Themes

Nuori tutkijasukupolvi 2019

Identified topics

bioinformatics