利用矩阵分解提取生物医学文献中潜在相关基因

Extracting Potential Related Genes from the Biomedical Literature via Matrix Factorization

利用Bicomb从50个基因文档中提取主题词建立词-基因矩阵,用对数权重法对词-基因矩阵进行处理,采用非负矩阵分解和奇异值分解法对词-基因矩阵降维,通过计算余弦相似度推断基因关系。结果表明,利用矩阵分解可以从生物医学文献中提取潜在相关基因。

The paper introduces by using Bicomb clustering 50 subjects to construct term - gene matrix, then using logarithm weigh- ting method to process, Non -negative Matrix Factorization （NMF） and Singular Value Decomposition （SVD） method to degradate, by calculating the cosine similarity to infer gene relationships. The result indicates extracting potential related genes from the biomedical lit- erature via matrix factorization .