python - calculate document weight using machine learning -

lets have n number of documents(resumes) in list, , want weigh each document(resume) of same category job description.txt reference. want weigh document per below. question there other approach weigh document in kind of scenario? in advance.

plan of action :

a) resumes (eg. 10) related same category (eg. java)

b) bag of words docs

for:

c) each document features names using tfidf vectorizor scores  d) have list of featured words in list   e) compare these features in "job discription" bag of words  f) count score document adding columns , weigh document 

what understood question looking grade resumes(documents) seeing how similar job description document. 1 approach can used convert documents tfidf matrix including job description. each document can seen vector in word space. once have created tfidf matrix, can calculate similarity between 2 documents using cosine similarity.

there additional things should removing stopwords, lemmatizing , encoding. additionaly may want make use of n-grams.

you can refer this book more information.

edit:

adding setup code

import numpy np sklearn.feature_extraction.text import tfidfvectorizer sklearn.metrics.pairwise import cosine_similarity nltk.corpus import stopwords import string import spacy nlp = spacy.load('en')  # remove punctuations translator = str.maketrans('', '', string.punctuation)  # sample documents resumes = ["executive administrative assistant on 10 years of experience providing thorough , skillful support senior executives.", "experienced administrative assistant, successful in project management , systems administration.", "10 years of administrative experience in educational settings; particular skill in establishing rapport people diverse backgrounds.", "ten years administrative support professional in corporation provides confidential case work.", "a highly organized , detail-oriented executive assistant on 15 years' experience providing thorough , skillful administrative support senior executives.", "more 20 years knowledgeable , effective psychologist working individuals, groups, , facilities, particular emphasis on geriatrics , multiple psychopathologies within population.", "ten years sales professional management experience in fashion industry.", "more 6 years librarian, 15 years' experience active participant in school-related events , support organizations.", "energetic sales professional knack matching customers optimal products , services meet specific needs. consistently received excellent feedback customers.", "more 6 years of senior software engineering experience, strong analytical skills , broad range of computer expertise.", "software developer/programmer history of productivity , successful project outcomes."]  job_doc = ["""executive administrative knack matching , effective psychologist particular emphasis on geriatrics"""]  # combine 2 _all = resumes+job_doc  # convert each spacy document docs= [nlp(document) document in _all]  # lemmatizae words, remove stopwords, remove punctuations docs_pp = [' '.join([token.lemma_.translate(translator) token in docs if not token.is_stop]) docs in docs]  # tfidf matrix tfidf_vec = tfidfvectorizer() tfidf_matrix = tfidf_vec.fit_transform(docs_pp).todense()  # calculate similarity cosine_similarity(tfidf_matrix[-1,], tfidf_matrix[:-1,]) 

wiki