Method for constructing variation literature interpretation knowledge base, and interpretation method and electronic device

A Natural Language Processing (NLP)-based variant literature interpretation method, the method comprising:

• constructing a variant literature interpretation knowledge base;
• constructing a machine automatic literature reading system using the variant literature interpretation knowledge base, wherein the machine automatic literature reading system comprises a human-computer interaction interface;
• obtaining an entity name to be interpreted;
• inputting, via the human-computer interaction interface of the machine automatic literature reading system, the entity name into the variant literature interpretation knowledge base to obtain an evidence criterion or evidence type, an evidence sentence, and an evidence word corresponding to the entity name;
• and
• returning, via the human-computer interaction interface of the machine automatic literature reading system, the evidence criterion or evidence type, the evidence sentence, and the evidence word corresponding to the entity name to an interpreter, enabling the interpreter to improve the quality and efficiency of interpretation of gene detection report,
• wherein said constructing the variant literature interpretation knowledge base comprises:
• obtaining disease-related literature;
• constructing, based on the disease-related literature, a database of entities associated with genes and variants;
• constructing a literature evidence knowledge graph for variant interpretation;
• and
• performing evidence extraction on the literature evidence knowledge graph to obtain evidence corresponding to an entity, and constructing, based on the evidence and the database, the variant literature interpretation knowledge base,
• wherein said constructing, based on the disease-related literature, the database of entities associated with genes and variants comprises:
• constructing an entity extraction model using some literature of the disease-related literature;
• performing, through the entity extraction model, entity extraction on the remaining literature in the disease-related literature to obtain an entity name;
• constructing an entity alignment model;
• performing, through the entity alignment model, entity alignment on the entity name to obtain an entity standard term corresponding to the entity name;
• and
• constructing, based on the entity name and the entity standard term corresponding to the entity name, the database of entities associated with genes and variants.