Notes

from transformers import BertTokenizerFast, BertForTokenClassification
import torch

# Load the trained model and tokenizer
model_path = './job_position_extraction_model' # Update this to your model's path
tokenizer = BertTokenizerFast.from_pretrained(model_path)
model = BertForTokenClassification.from_pretrained(model_path)

# Function to predict and extract job positions
def predict_job_positions(text):
# Tokenize input text
inputs = tokenizer.encode_plus(text, return_tensors="pt", add_special_tokens=True, max_length=512, padding='max_length', truncation=True)
input_ids = inputs["input_ids"]
attention_mask = inputs["attention_mask"]

# Get model predictions
with torch.no_grad():
outputs = model(input_ids, attention_mask=attention_mask)
logits = outputs.logits

# Decode predictions
predictions = torch.argmax(logits, dim=2)

# Extract and print job positions
tokens = tokenizer.convert_ids_to_tokens(input_ids[0])
label_indices = predictions[0].numpy()

# Assuming label_dict is as before
label_dict = {0: 'O', 1: 'B-Position', 2: 'I-Position'}
reverse_label_map = {v: k for k, v in label_dict.items()}

print("Detected Job Positions:")
for token, label_index in zip(tokens, label_indices):
if label_index != reverse_label_map['O']: # Ignore 'O' labels
print(f"{token} ({label_dict[label_index]})")

# Example usage
example_text = "We are looking for a Data Scientist and a Software Engineer to join our team."
predict_job_positions(example_text)