Semantic indexing of biomedical literature is usually done at the level of MeSH descriptors with several related but distinct biomedical concepts often grouped together and treated as a single topic. For instance, several disease-types are grouped together impeding the targeted retrieval of relevant information. Subject annotations at the level of MeSH concepts, though beneficial for accessing and integrating biomedical knowledge, are beyond the current state. In this study, we investigate the automated refinement of existing coarse-grained/generic topic annotations of biomedical articles, at the level of MeSH descriptors, into fine-grained/specific topic annotations at the level of concepts. In this direction, we explore the specific challenges of state-of-the-approaches on weakly supervised and zero-shot text classification, and investigate machine-learning techniques for addressing them.
Relevant literature:
Fine-grained semantic indexing of biomedical literature based on weak supervision.
Large-scale investigation of weakly-supervised deep learning for the fine-grained semantic indexing of biomedical literature
Zero-Shot Relabeling of Weak Labels for Fine-Grained Semantic Indexing of Biomedical Literature