Computational Literature-based Discovery Methods
Auf einen Blick
Literature-based discovery (LBD) is a research field aiming at discovering new, implicit knowledge from the literature. Its basic assumption is that two concepts A and C that do not occur in the same article can be connected via some other terms B, which can imply a new meaningful relation between A and C.
In this proposal we focus on computational LBD in a biomedical context. Current LBD systems use knowledge bases that store relations extracted from scientific literature. The most frequently used SemMedDB, however, has limitations: use of rule-based methods; use of titles and abstracts only; limited coverage of entity and relation types. We aim to:
- Develop new algorithms to automatically extract relations, using state-of-the-art neural networks/deep learning methods, and processing full texts to obtain a new knowledge base;
- Apply our methods to the domain of Natural Product Chemistry. Indeed, natural products have successfully supported the discovery of pharmaceuticals, but the discovery of medically relevant products is still performed manually.
Sojka, Petr; Horák, Aleš; Kopeček, Ivan; Pala, Karel, Hrsg.,
Text, Speech, and Dialogue.
25th International Conference on Text, Speech and Dialogue (TSD), Brno, Czech Republic, 6-9 September 2022.
Lecture Notes in Computer Science ; 13502.
Verfügbar unter: https://doi.org/10.1007/978-3-031-16270-1_14