SpanMarker for Named Entity Recognition

SpanMarker is a framework for training powerful Named Entity Recognition models using familiar encoders such as BERT, RoBERTa and DeBERTa. Tightly implemented on top of the 🤗 Transformers library, SpanMarker can take advantage of its valuable functionality.

Based on the PL-Marker paper, SpanMarker breaks the mold through its accessibility and ease of use. Crucially, SpanMarker works out of the box with many common encoders such as bert-base-cased and roberta-large, and automatically works with datasets using the IOB, IOB2, BIOES, BILOU or no label annotation scheme.

Installation

You may install the span_marker Python module via pip like so:

pip install span_marker

Quick Start

Please have a look at our Getting Started jupyter notebook for details on how SpanMarker is commonly used. That notebook explains the following snippet in more detail.

from datasets import load_dataset
from span_marker import SpanMarkerModel, Trainer
from transformers import TrainingArguments

dataset = load_dataset("DFKI-SLT/few-nerd", "supervised")
labels = dataset["train"].features["ner_tags"].feature.names

model_name = "bert-base-cased"
model = SpanMarkerModel.from_pretrained(model_name, labels=labels)

args = TrainingArguments(
    output_dir="my_span_marker_model",
    learning_rate=5e-5,
    gradient_accumulation_steps=2,
    per_device_train_batch_size=4,
    per_device_eval_batch_size=4,
    num_train_epochs=1,
    save_strategy="steps",
    eval_steps=200,
    logging_steps=50,
    bf16=True,
    warmup_ratio=0.1,
)

trainer = Trainer(
    model=model,
    args=args,
    train_dataset=dataset["train"].select(range(8000)),
    eval_dataset=dataset["validation"].select(range(2000)),
)

trainer.train()
trainer.save_model("my_span_marker_model/checkpoint-final")

metrics = trainer.evaluate()
print(metrics)

For this work is based on PL-Marker, you may expect similar results to its Papers with Code Leaderboard. Tests, documentation and further information on expected performance will come soon.