I have followed this TF IDF implementation to match a keyword list with many Resumes in text format. It works pretty well : https://towardsdatascience.com/build-a-text-recommendation-system-with-python-e8b95d9f251c
But when I search the keywords "Active directory", I get way too much architects when I want to find System Administrator in place : since Architects have all matching keywords, they pop up in the top results. I have added "admin administrator". I have repeated those keywords and it helps.
How apply a negative weight on some words in this algorithm please ? This way, if someone have the negative keyword, he won't be completly exclude, but he will have a penality in the classement. How can I do please ?
I don't want to recalculate all words weight for all resume each time I execute a search. So I was thinking : let's calculate the weight of each words on all resume for once. They will only have postive weight. And let's add negative weight only on some search keywords. Then compare the search vector with all Resume vector.
How can I do please ?