Deep Learning approach to find semantic similarity between taxonomy and documents

I have some 100000 documents and I have a taxonomy for each category (Agriculture, Education, Sports, Technology, etc…)

I would like to score those documents based on semantic similarity with taxonomy?

Ex: Internet, mobile, Laptop, and some other related terms should give a high similarity score with Technology Category Documents and less similarity score with other categories like agriculture, Education, etc…

I tired with keywords, synonyms, regex matching, etc …are there better approaches with deep learning