Licensable or Stays Out Classifier
Reads a description of a kind of company record or limit and says whether it describes something that can be considered for licensing (can_be_considered) or something that stays out (stays_out).
All training text is our own, from pages published at aitrainingdatabroker.com. Nothing comes from a customer.
Related tool: company-data-licensability-explorer, company-record-types-and-limits-finder (a Space that runs in your browser).
Why it exists
When a team lists what a company holds, each line is either a candidate or a limit. This model gives a first sort of that list so a person reviews the limits first. It does not know a company's contracts or the law that applies.
Use it
from transformers import pipeline
classify = pipeline("text-classification", model="aitrainingdatabroker/licensable-or-stays-out-classifier", top_k=3)
print(classify("Customer payment card numbers and billing addresses held in the support system."))
How it was trained
Fine-tuned from distilroberta-base for 6 epochs on one RTX 3070. Training data: 224 passages from company-data-licensability-by-industry, company-record-types-and-limits, which are our own published page text. It was trained on item descriptions from our industry pages and tested on item descriptions from our record-type pages, a different subject than the one it learned from. The scripts that built the data and trained the model are in source/; the full scores are in scores.json.
Test scores
On 274 held-out rows: accuracy 0.956, macro F1 0.954.
| Label | F1 | Count |
|---|---|---|
can_be_considered |
0.963 | 166 |
stays_out |
0.946 | 108 |
Limits
- Binary labels hide shades: a record can be licensable only in part, or only after de-identification.
- It reacts to how our pages are worded. Check its output on your own list before relying on it.
- This model is not legal advice and not a certification. A person should review its output before anything is shared or decided, and your counsel decides what a rule requires.
Read more
- How selling company data to AI labs works
- Who is liable when data is licensed to AI
- Record types and export guides
License and credit
Apache 2.0. Please credit AI Training Data Broker and link to https://aitrainingdatabroker.com/ when you share or build on it.
About
AI Training Data Broker is a broker and technical advisor for companies that sell or license their data to AI labs. It guides the technical side, including export, format, de-identification and compliance, and publishes free guides, checklists and tools.