ASK system

There is a vast amount of data in the biomedical field and these knowledge are widely documented in a large number of publications, such as papers, reports, and research results written and published in text formats, which provide numerous opportunities for knowledge acquisition and expansion. However, biomedical data is huge and continues to evolve with new advances in biology and medicine.This makes literature search very cumbersome, and doctors and researchers in medical field cannot easily extract and understand new knowledge from the massive literature timely.Using knowledge graphs can extract knowledge faster. Based on the requirement, we propose an automatic construction and self-reflection framework (ASK) that can extract, merge, discover, and update knowledge for biomedical knowledge graphs (BKGs). The main contents of the framework are as follows:

Collect the latest updated data from the Pubmed database 24 hours a day, and initially build a graph through entity recognition and relationship extraction.
Use conflict resolution and fusion to integrate the knowledge of the constructed graph with the previous graph.
Use self-reflection strategies to predict new knowledge that may be proven in the future.

So far, we have constructed various types of knowledge graph including cancer, skin disease, and mental disease, and demonstrated their practicality. You can follow future progress on our laboratory’s official website.