Model Card for TickerNER
This is a Named-Entity-Recognition model aiming at detecting ticker symbols in an input query. It has been fine-tuned from DistilBERT, and it has roughly 67M parameters. It is english-speaking only, but it can recognize ticker symbols for U.S companies (XXXX), European companies (XXX.XX) and Asian companies (NNNN.XX).
Training Details
It has been trained using HuggingFace's Trainer library, on the custom FMPQueriesTickers dataset. Training occurred on a Nvidia A100 GPU provided by HuggingFace's ZeroGPU Spaces, it only lasts a few seconds. Below are its evaluation metrics on the validation dataset.
precision
|0.945824
recall
|0.962113
f1
|0.953899
accuracy
|0.993940
- Downloads last month
- 40