Medical achievement recommendation method and system based on entity relation mapping

Abstract

The invention provides a medical achievement recommendation method based on entity relation mapping, and belongs to the technical field of research achievement recommendation, and the scheme comprises the following steps: obtaining a pre-constructed medical knowledge graph; based on a pre-trained knowledge representation model, performing vectorization representation on entities in the medical knowledge graph; wherein the knowledge representation model is constructed based on an entity relation mapping matrix and multi-mode depth embedding, the entity relation mapping matrix is generated through projection vectors, and a relation mapping matrix in a TransR model is replaced; meanwhile, a multi-mode deep embedding concept is introduced, and a reverse translation geometric distance embedding model and a symmetric relation embedding model are added into entity and relation modeling; and based on the obtained entity vectorization representation, determining the similarity between the entities so as to obtain the research result recommendation of the disease concerned by the user.

Description

technical field [0001] The present disclosure belongs to the technical field of research achievement recommendation, and in particular relates to a method and system for recommending medical achievements based on entity relationship mapping. Background technique [0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art. [0003] In the medical field, knowledge graph has a very wide range of applications, and medical knowledge graph is of great significance in clinical decision support, medical intelligent semantic retrieval, medical question answering and other fields. Social media has the potential to transform public health, including disseminating health updates, sharing disease information, and more. The use of social media to disseminate the latest research results of related diseases, so that patients can obtain the latest research progress, will be very beneficial for pa...

AI Technical Summary

Problems solved by technology

[0005] The inventor found that although the TransE model is simple and effective, it has limitations in dealing with one-to-many, many-to-one, and many-to-many relationships, and cannot effectively distinguish different entities with the same relationship. The improved TransR model based on TransE solves the above problems. problem, but the following problems still exist:

[0006] (1) Using distance as a scoring metric makes the accuracy of knowledge representation be affected by irrelevant dimensions. Secondly, it is no longer effective in learning the inversion relational patterns that TransE can handle initially, and it is impossible to simultaneously perform inversion, symmetry, antisymmetry, etc. Relational schema for modeling and reasoning

[0007] (2) The translation model mostly uses the gap-based sorting error function as the optimization objective function of the training model. This optimization objective function will minimize the sum of the score errors between the negative triplet and the positive triplet, so it is applied to Positive triplet scores may not be small enough to preserve the relationship of the score function when translating the model

Images