title_value

title_value

title_value is a fine-tuned sequence classification model adapted for value extraction (regression) on job posting titles. In particular, this model maps job title strings to a single, real-valued numerical score reflecting the structural/hierarchical seniority of the job role.

Basic Usage

To use this model for a regression task via the Hugging Face pipeline API, you must set function_to_apply="none". This ensures the raw numerical logit output is returned rather than passing through an activation function like softmax or sigmoid.

from transformers import pipeline, AutoModelForSequenceClassification, AutoTokenizer

model_name = "loyoladatamining/title_value"

# Explicitly load as a regression model with 1 output label
model = AutoModelForSequenceClassification.from_pretrained(
    model_name, 
    num_labels=1, 
    problem_type="regression"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

# Create text classification pipeline
nlp = pipeline(
    "text-classification", 
    model=model, 
    tokenizer=tokenizer, 
    function_to_apply="none"
)

# Inference
title = "Senior Cloud Infrastructure Architect"
result = nlp(title)
print(result)

Output Format

The pipeline returns a single classification dictionary containing the regression score:

[
  {
    "label": "LABEL_0",
    "score": 2.172
  }
]

Citation

If you find this model useful in your work, please consider citing:

@article{meisenbacher2025extracting,
  title={Extracting O* NET Features from the NLx Corpus to Build Public Use Aggregate Labor Market Data},
  author={Meisenbacher, Stephen and Nestorov, Svetlozar and Norlander, Peter},
  year={2025}
}