main.py

from sentence_transformers import SentenceTransformer, util
from flask import Flask, request
import json
import logging
class JobOffer:
    def __init__(self, title, description, skills):
        self.title = title
        self.description = description
        self.skills = skills

class JobApplicant:
    def __init__(self, id, skills):
        self.id = id
        self.skills = skills
        self.score = 0

class JobApplicantEncoder(json.JSONEncoder):
    def default(self, o):
        return o.__dict__

app = Flask(__name__)

model_name = "Sahajtomar/french_semantic"
model_name = "dangvantuan/sentence-camembert-base"
model_name = "dangvantuan/sentence-camembert-large"
model = SentenceTransformer(model_name)


def get_cosine_score(job_applicant, job_offer):
    embeddings1 = model.encode(job_applicant, convert_to_tensor=True)
    embeddings2 = model.encode(job_offer, convert_to_tensor=True)
    cosine_scores = util.pytorch_cos_sim(embeddings1, embeddings2)
    return cosine_scores[0][0].item()

@app.route('/process/', methods=['POST'])
def get_cosinos():
    #Map the input strings to object.
    data = request.get_json()
    temp_job_offer = data["job_offer"]
    temp_job_applicant = data["job_applicant"]
    assert temp_job_offer is not None
    assert temp_job_applicant is not None
    job_offer = JobOffer(temp_job_offer["title"], temp_job_offer["description"], temp_job_offer["skills"])
    job_applicant_list = []
    for job_applicant in temp_job_applicant:
        job_applicant_list.append(JobApplicant(job_applicant["id"], job_applicant["skills"]))

    #Compute the cosine score between the job offer and the job applicant
    for job_applicant in job_applicant_list:
        job_applicant.score = get_cosine_score(job_applicant.skills, job_offer.skills)
    
    return json.dumps(job_applicant_list, cls=JobApplicantEncoder)


# embeddings1 = model.encode(job_applicant, convert_to_tensor=True)
# embeddings2 = model.encode(job_offer, convert_to_tensor=True)
# cosine_scores = util.pytorch_cos_sim(embeddings1, embeddings2)

# # Create a list of tuples containing sentence and score
# results = []
# for i in range(len(job_applicant)):
#     for j in range(len(job_offer)):
#         results.append((job_applicant[i], job_offer[j], cosine_scores[i][j]))

# # Sort the list based on the scores in descending order
# results.sort(key=lambda x: x[2], reverse=True)

# # Print the sentences in the sorted order
# for result in results:
#     print("Sentence 1:", result[0])
#     print("Sentence 2:", result[1])
#     print("Similarité cosinus:", result[2])
#     print("-" * 50)

if __name__ == '__main__':
    logging.info("Starting server...")
    resp = get_cosine_score("Je ne suis pas un développeur web", "Nous recherchons un développeur web qui sache faire du React, du NodeJS et du MongoDB")
    print(f"0 : {resp}")
    resp = get_cosine_score("Je ne suis pas un développeur web, j'ai aucune compétences en React", "Nous recherchons un développeur web qui sache faire du React, du NodeJS et du MongoDB")
    print(f"1 : {resp}")
    resp = get_cosine_score("Je ne suis pas un développeur web, j'ai aucune compétences en React et MongoDB", "Nous recherchons un développeur web qui sache faire du React, du NodeJS et du MongoDB")
    print(f"2 : {resp}")
    resp = get_cosine_score("Je ne suis pas un développeur web, j'ai aucune compétences en React, MongoDB et NodeJS", "Nous recherchons un développeur web qui sache faire du React, du NodeJS et du MongoDB")
    print(f"3 : {resp}")
    # app.run(host='0.0.0.0', port=5000, debug=True)