Spaces:
Sleeping
Sleeping
File size: 3,566 Bytes
2d38c97 92feaec 2d38c97 65281cf d8eff68 2d38c97 d8eff68 2d38c97 65281cf 2d38c97 aba2c4a 2d38c97 92feaec 2d38c97 65281cf 2d38c97 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 | import numpy as np
from sklearn.metrics.pairwise import cosine_similarity
from src.feature_engg.tfidf_vectorizing_data import load_tfidf_matrix
def compute_similarity_matrix(X_resumes, X_jobs ):
"""
Compute cosine similarity between resume vectors and job vectors.
Returns a similarity matrix of shape (num_resumes, num_jobs).
"""
return cosine_similarity(X_resumes, X_jobs)
def top_n_tfidf_matches(similarity_matrix: np.ndarray,
top_n: int = 5,
job_df = None):
"""
For each resume, get indices of top N matching jobs.
Args:
similarity_matrix: np.ndarray -> Cosine similarity matrix of shape (num_resumes, num_jobs)
top_n: int -> Number of top matches to return for each resume.
Returns:
Dict[int, List[Tuple[job_idx, score]]] -> top job matches per resume
"""
results = {}
for i, row in enumerate(similarity_matrix):
seen_titles = set()
ranked = []
for j in row.argsort()[::-1]: # Sorted indices in descending order
title = job_df.iloc[j]['title'] if job_df is not None else j
if title not in seen_titles:
ranked.append((j, round(row[j], 4)))
seen_titles.add(title)
if len(ranked) == top_n:
break
results[i] = ranked
return results
def top_n_bert_matches(indices, distances, job_df, top_n=5):
"""
Deduplicate FAISS results by job title and return top-N unique matches.
Searches across all jobs if provided.
Args:
indices (np.ndarray): Indices of nearest neighbors from FAISS (shape: [1, k]).
distances (np.ndarray): Distances/similarities from FAISS (shape: [1, k]).
job_df (pd.DataFrame): DataFrame containing job titles.
top_n (int): Number of unique top matches to return.
Returns:
List[Tuple[int, float]]: List of (job_idx, score) for top-N unique titles.
"""
seen_titles = set()
ranked = []
for idx, score in zip(indices[0], distances[0]):
title = job_df.iloc[idx]['title']
if title not in seen_titles:
ranked.append((idx, float(score)))
seen_titles.add(title)
if len(ranked) == top_n:
break
return ranked
if __name__ == "__main__":
# Define paths
resume_vec_path = "models/dev_tfidf/resumes_tfidf_matrix.npz"
job_vec_path = "models/dev_tfidf/jobs_tfidf_matrix.npz"
# Load sparse TF-IDF matrices
X_resumes = load_tfidf_matrix(resume_vec_path)
X_jobs = load_tfidf_matrix(job_vec_path)
print(f"✅ Loaded resumes vector shape: {X_resumes.shape}")
print(f"✅ Loaded job descriptions vector shape: {X_jobs.shape}")
# Compute cosine similarity
similarity_matrix = compute_similarity_matrix(X_resumes, X_jobs)
# # Uncomment to see similarity statistics like min, max, mean, median similarity scores
# all_scores = similarity_matrix.flatten()
# print(f"Min score: {np.min(all_scores):0.4f}, \nMax score: {np.max(all_scores):0.4f}, \nMean score: {np.mean(all_scores):0.4f}, \nMedian score: {np.median(all_scores):0.4f}")
# Get top 5 matches per resume
matches = top_n_tfidf_matches(similarity_matrix, top_n=5)
# Display example output (i.e. top_n job matches for first 5 resumes)
for resume_idx, top_jobs in list(matches.items())[:5]:
print(f"\n🔎 Resume {resume_idx} best matches:")
for job_idx, score in top_jobs:
print(f" ↳ Job {job_idx} with similarity score: {score}")
|