Update README.md
Browse files
README.md
CHANGED
|
@@ -3,84 +3,247 @@ library_name: transformers
|
|
| 3 |
license: apache-2.0
|
| 4 |
base_model: answerdotai/ModernBERT-base
|
| 5 |
tags:
|
| 6 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
model-index:
|
| 8 |
-
- name:
|
| 9 |
-
results:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
license: apache-2.0
|
| 4 |
base_model: answerdotai/ModernBERT-base
|
| 5 |
tags:
|
| 6 |
+
- actuarial
|
| 7 |
+
- insurance
|
| 8 |
+
- multilabel-classification
|
| 9 |
+
- sentence-classification
|
| 10 |
+
- skills-extraction
|
| 11 |
+
- career-planning
|
| 12 |
+
- modernbert
|
| 13 |
+
- job-analysis
|
| 14 |
+
datasets:
|
| 15 |
+
- actuarial-jobs-7k
|
| 16 |
+
language:
|
| 17 |
+
- en
|
| 18 |
+
metrics:
|
| 19 |
+
- f1
|
| 20 |
+
- precision
|
| 21 |
+
- recall
|
| 22 |
model-index:
|
| 23 |
+
- name: modernbert-actuarial-skills-classifier
|
| 24 |
+
results:
|
| 25 |
+
- task:
|
| 26 |
+
type: text-classification
|
| 27 |
+
name: Multi-Label Text Classification
|
| 28 |
+
metrics:
|
| 29 |
+
- type: f1_micro
|
| 30 |
+
value: 0.6728
|
| 31 |
+
name: F1 Micro
|
| 32 |
+
- type: f1_macro
|
| 33 |
+
value: 0.1060
|
| 34 |
+
name: F1 Macro
|
| 35 |
+
- type: precision_micro
|
| 36 |
+
value: 0.7915
|
| 37 |
+
name: Precision Micro
|
| 38 |
+
- type: recall_micro
|
| 39 |
+
value: 0.5850
|
| 40 |
+
name: Recall Micro
|
| 41 |
+
widget:
|
| 42 |
+
- text: "I am looking for an entry-level actuarial position in life insurance pricing where I can apply my knowledge of mortality tables and statistical analysis. I have strong Python programming skills and experience with GLM models from my university projects. I am particularly interested in learning more about IFRS 17 implementation and would like to work with modern actuarial software like Prophet or MoSes."
|
| 43 |
+
example_title: "Life Insurance Entry Level"
|
| 44 |
+
- text: "I have three years of experience as a reserving actuary in property and casualty insurance, working primarily with workers compensation and general liability lines. I am proficient in R and SQL for data analysis and have built several predictive models using machine learning techniques. I am now seeking a senior analyst role where I can lead pricing projects and mentor junior actuaries, with a target salary range of at least 85000 dollars annually."
|
| 45 |
+
example_title: "P&C Career Growth"
|
| 46 |
+
- text: "After completing my actuarial exams up to ASA level, I want to transition into a health insurance role focusing on medical cost trend analysis and risk adjustment. I enjoy working with large datasets and have self-taught Python and SAS for healthcare analytics. My ideal position would involve building pricing models for group health products and I am hoping to find opportunities that offer around 70000 dollars per year as I build my specialization in this area."
|
| 47 |
+
example_title: "Health Insurance Transition"
|
| 48 |
+
- text: "I am a recent mathematics graduate passionate about pension actuarial work and retirement planning. I have limited professional experience but completed internships where I learned about defined benefit schemes, asset liability management, and regulatory compliance under Solvency II. I am eager to develop my Excel and VBA skills further and would consider positions starting at 40000 dollars minimum while I continue studying for my actuarial fellowship exams."
|
| 49 |
+
example_title: "Pensions Graduate Role"
|
| 50 |
+
- text: "As a data scientist looking to move into the actuarial field, I bring extensive experience with machine learning frameworks like TensorFlow and PyTorch, as well as strong programming abilities in Python and Scala. I am particularly interested in applying deep learning techniques to mortality forecasting and longevity risk modeling in life insurance. I am seeking roles that value innovation in actuarial modeling and offer competitive compensation of at least 95000 dollars given my technical background."
|
| 51 |
+
example_title: "Data Science to Actuarial"
|
| 52 |
+
inference:
|
| 53 |
+
parameters:
|
| 54 |
+
threshold: 0.5
|
| 55 |
+
top_k: 15
|
| 56 |
---
|
| 57 |
|
| 58 |
+
<div align="center">
|
| 59 |
+
<img src="./photo.png" width="150" height="150" style="border-radius: 50%; margin: 20px 0;">
|
| 60 |
+
|
| 61 |
+
# 👋 Connect with me on LinkedIn!
|
| 62 |
+
|
| 63 |
+
[](https://www.linkedin.com/in/manuel-caccone-42872141/)
|
| 64 |
+
|
| 65 |
+
**Manuel Caccone - Actuarial Data Scientist & Open Source Educator**
|
| 66 |
+
|
| 67 |
+
*Let's discuss actuarial science, AI, and career development!*
|
| 68 |
+
|
| 69 |
+
---
|
| 70 |
+
|
| 71 |
+
</div>
|
| 72 |
+
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
# 🎯 ModernBERT Actuarial Skills Classifier: Your Career Planning Assistant
|
| 76 |
+
|
| 77 |
+
---
|
| 78 |
+
|
| 79 |
+
## 🚩 Model Description
|
| 80 |
+
|
| 81 |
+
**ModernBERT-actuarial-skills-classifier** is a fine-tuned ModernBERT-base model trained on over 7,000 actuarial job postings, purpose-built to extract and identify actuarial competencies and technical skills from natural language descriptions. It powers career planning, skills gap analysis, and learning roadmap generation for actuarial professionals and students.
|
| 82 |
+
|
| 83 |
+
---
|
| 84 |
+
|
| 85 |
+
## ✨ Key Features
|
| 86 |
+
|
| 87 |
+
- 🎯 **Multi-Label Classification:** Identifies multiple relevant skills from a single description
|
| 88 |
+
- 📚 **Career-Focused:** Trained on real job postings covering Life, P&C, Health, and Pensions
|
| 89 |
+
- 🚀 **Instant Analysis:** Get results in under 1 second
|
| 90 |
+
- 🔓 **Open Source:** Apache 2.0 License for educational and commercial use
|
| 91 |
+
- 🌐 **Interactive Demo:** Full-featured Gradio Space with learning roadmaps and batch processing
|
| 92 |
+
|
| 93 |
+
---
|
| 94 |
+
|
| 95 |
+
## 💡 Intended Use Cases
|
| 96 |
+
|
| 97 |
+
- **Career Planning:** Students and early-career actuaries discovering required skills for target roles
|
| 98 |
+
- **Job Analysis:** Extracting structured skill requirements from job descriptions
|
| 99 |
+
- **Skills Gap Assessment:** Identifying learning priorities when changing specializations
|
| 100 |
+
- **Market Research:** Analyzing trends in actuarial job requirements across industries
|
| 101 |
+
- **Resume Optimization:** Matching your background to employer expectations
|
| 102 |
+
|
| 103 |
+
### Examples
|
| 104 |
+
|
| 105 |
+
```
|
| 106 |
+
Input: "I am looking for an entry-level actuarial position in life insurance pricing
|
| 107 |
+
where I can apply my knowledge of mortality tables and statistical analysis. I have
|
| 108 |
+
strong Python programming skills and experience with GLM models from my university
|
| 109 |
+
projects. I am particularly interested in learning more about IFRS 17 implementation."
|
| 110 |
+
|
| 111 |
+
Output: Life Insurance Pricing (92%), Python (88%), GLM Modeling (85%),
|
| 112 |
+
Statistical Analysis (82%), Mortality Tables (78%), IFRS 17 (75%),
|
| 113 |
+
Entry Level (71%), Excel (68%)...
|
| 114 |
+
```
|
| 115 |
+
|
| 116 |
+
---
|
| 117 |
+
|
| 118 |
+
## 📂 Training Data
|
| 119 |
+
|
| 120 |
+
- **Dataset Size:** 7,000+ real actuarial job postings
|
| 121 |
+
- **Time Period:** 2023-2025 job market
|
| 122 |
+
- **Coverage:** Life, P&C, Health, Pensions, Reinsurance, Consulting
|
| 123 |
+
- **Labels:** 100+ unique skills covering actuarial domains, programming, tools, certifications, and soft skills
|
| 124 |
+
- **Format:** Multi-label classification with manual validation by actuarial professionals
|
| 125 |
+
|
| 126 |
+
---
|
| 127 |
+
|
| 128 |
+
## 📊 Training Statistics
|
| 129 |
+
|
| 130 |
+
| Metric | Value | Notes |
|
| 131 |
+
|--------------------|----------------|--------------------------------------------|
|
| 132 |
+
| Epochs | 10 | Best model at epoch 7 |
|
| 133 |
+
| Final F1 Micro | 0.6728 | Overall performance across all skills |
|
| 134 |
+
| Final F1 Macro | 0.1060 | Average per skill (handles class imbalance)|
|
| 135 |
+
| Precision Micro | 0.7915 | 79% of predictions are correct |
|
| 136 |
+
| Recall Micro | 0.5850 | Captures 58% of relevant skills |
|
| 137 |
+
| Hamming Loss | 0.0207 | Only 2% label error rate |
|
| 138 |
+
| Training Loss | 0.0602 | Final validation loss |
|
| 139 |
+
| Learning Rate | 2e-5 | With 10% warmup |
|
| 140 |
+
| Batch Size | 16 | Effective (8 per device, 2 grad accum) |
|
| 141 |
+
| Hardware | GPU | Mixed precision training (FP16) |
|
| 142 |
+
|
| 143 |
+
---
|
| 144 |
+
|
| 145 |
+
## 🛠️ Dependencies
|
| 146 |
+
|
| 147 |
+
```
|
| 148 |
+
transformers>=4.44.0
|
| 149 |
+
torch>=2.0.0
|
| 150 |
+
pandas
|
| 151 |
+
numpy
|
| 152 |
+
```
|
| 153 |
+
|
| 154 |
+
---
|
| 155 |
+
|
| 156 |
+
## ⚠️ Limitations & Ethics
|
| 157 |
+
|
| 158 |
+
- **Domain-Specific:** Optimized for actuarial and insurance contexts only
|
| 159 |
+
- **English Only:** Trained exclusively on English job postings
|
| 160 |
+
- **Class Imbalance:** Rare skills may have lower prediction confidence
|
| 161 |
+
- **Not Exhaustive:** Cannot predict skills not present in training data
|
| 162 |
+
- **Career Guidance Only:** Not a substitute for professional career counseling
|
| 163 |
+
- **Geographic Bias:** Primarily reflects US, UK, and EU job markets
|
| 164 |
+
|
| 165 |
+
---
|
| 166 |
+
|
| 167 |
+
## 💻 Usage Example
|
| 168 |
+
|
| 169 |
+
```python
|
| 170 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
| 171 |
+
import torch
|
| 172 |
+
|
| 173 |
+
# Load model
|
| 174 |
+
model_name = "manuelcaccone/modernbert-actuarial-skills-classifier"
|
| 175 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
| 176 |
+
model = AutoModelForSequenceClassification.from_pretrained(model_name)
|
| 177 |
+
|
| 178 |
+
# Prepare text
|
| 179 |
+
text = """I am a recent mathematics graduate passionate about pension actuarial work
|
| 180 |
+
and retirement planning. I have limited professional experience but completed internships
|
| 181 |
+
where I learned about defined benefit schemes and regulatory compliance. I am eager to
|
| 182 |
+
develop my Excel skills further and would consider positions starting at 40000 dollars
|
| 183 |
+
minimum while I continue studying for my actuarial exams."""
|
| 184 |
+
|
| 185 |
+
# Tokenize and predict
|
| 186 |
+
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=512)
|
| 187 |
+
with torch.no_grad():
|
| 188 |
+
outputs = model(**inputs)
|
| 189 |
+
probabilities = torch.sigmoid(outputs.logits)
|
| 190 |
+
|
| 191 |
+
# Get predictions above threshold
|
| 192 |
+
threshold = 0.5
|
| 193 |
+
predicted_indices = torch.where(probabilities[0] > threshold)[0]
|
| 194 |
+
|
| 195 |
+
# Display results
|
| 196 |
+
print("Predicted Skills:")
|
| 197 |
+
for idx in predicted_indices:
|
| 198 |
+
skill = model.config.id2label[idx.item()]
|
| 199 |
+
confidence = probabilities[0][idx].item()
|
| 200 |
+
print(f" {skill}: {confidence:.1%}")
|
| 201 |
+
```
|
| 202 |
+
|
| 203 |
+
---
|
| 204 |
+
|
| 205 |
+
## 🌟 Related Resources
|
| 206 |
+
|
| 207 |
+
This model is part of an actuarial AI ecosystem:
|
| 208 |
+
|
| 209 |
+
- **Interactive Demo:** [Skills Classifier Space](https://huggingface.co/spaces/manuelcaccone/actuarial-skills-classifier)
|
| 210 |
+
- **Model Repository:** [modernbert-actuarial-skills-classifier](https://huggingface.co/manuelcaccone/modernbert-actuarial-skills-classifier)
|
| 211 |
+
- **Base Model:** [ModernBERT-base](https://huggingface.co/answerdotai/ModernBERT-base)
|
| 212 |
+
|
| 213 |
+
---
|
| 214 |
+
|
| 215 |
+
## 👤 Author & Citation
|
| 216 |
+
|
| 217 |
+
- **Creator:** Manuel Caccone (Actuarial Data Scientist & Open Source Educator)
|
| 218 |
+
- [LinkedIn](https://www.linkedin.com/in/manuel-caccone-42872141/) · [[email protected]](mailto:[email protected])
|
| 219 |
+
|
| 220 |
+
```bibtex
|
| 221 |
+
@model{caccone2025actuarialskills,
|
| 222 |
+
title={ModernBERT Actuarial Skills Classifier: Career Planning with Multi-Label Classification},
|
| 223 |
+
author={Caccone, Manuel},
|
| 224 |
+
year={2025},
|
| 225 |
+
publisher={Hugging Face},
|
| 226 |
+
url={https://huggingface.co/manuelcaccone/modernbert-actuarial-skills-classifier},
|
| 227 |
+
note={Fine-tuned ModernBERT for actuarial skills extraction from job descriptions}
|
| 228 |
+
}
|
| 229 |
+
```
|
| 230 |
+
|
| 231 |
+
---
|
| 232 |
+
|
| 233 |
+
## 📜 License
|
| 234 |
+
|
| 235 |
+
Apache 2.0 License — use, modify, and cite for ethical, research, educational, and commercial purposes.
|
| 236 |
+
|
| 237 |
+
---
|
| 238 |
+
|
| 239 |
+
<div align="center">
|
| 240 |
+
|
| 241 |
+
### 🤝 Want to collaborate or discuss actuarial AI?
|
| 242 |
+
|
| 243 |
+
[](https://www.linkedin.com/in/manuel-caccone-42872141/)
|
| 244 |
+
|
| 245 |
+
</div>
|
| 246 |
+
|
| 247 |
+
---
|
| 248 |
+
|
| 249 |
+
*Part of the actuarial open-source education initiative—bringing AI tools to the actuarial community!*
|