bert-base-NER

bert-base-NER

Finds and labels people, organizations, locations and miscellaneous names in English text. A fine-tuned BERT model for named entity recognition, used through the Transformers pipeline.

Made by dslim on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: bert-base-NER reads English text and marks the names in it. It tags four kinds of entity: people, organizations, locations and a miscellaneous group. It is a bert-base-cased model fine-tuned on the English CoNLL-2003 dataset.

Each word piece in the text gets a label, so you can see where a name begins and where it continues. The model has 108 million parameters.

Who it suits: Developers and data teams who need to pull names out of English text, such as news, reports or messages, and who are comfortable with a few lines of Python.

Industry | Who uses it | How they use it | Example
Media | Newsroom developers | Tag names in articles to build indexes | Listing the people and places in a news story
Data | Data engineers | Extract names from text before storing it | Pulling organization names from reports
Software | App developers | Add name detection to a text feature | Highlighting names in a notes app
Research | NLP students | Learn token classification on a standard dataset | Running a first named entity recognition test

Anyone who works in languages other than English, or on text far from news writing, can skip it or test it first, since the card warns it may not generalize.

Where to use it:

• Python with the Hugging Face Transformers library.
• Windows, macOS or Linux with Python, for downloading the files.
• Larger and smaller siblings by the same author, if you want a different size.

Use case | Who | What they get back
Name extraction | Developers | Each token labeled as a person, organization, location or other
Text indexing | Data teams | Entity tags to search or group documents by
Model comparison | Researchers | A baseline to compare other NER models against

How to set it up: You need Python and the Transformers library.

• Python on Windows, macOS or Linux
• The transformers library and a backend such as PyTorch
• English text to analyze
• A GPU is not required for a model of this size

1. Install the Hugging Face tool if you want the files locally.
2. Make sure Transformers is installed in your Python environment.
3. Run the pipeline example below.

pip install -U huggingface_hub
hf download dslim/bert-base-NER

How to use it: Load the tokenizer and model, build an ner pipeline, and pass it a sentence. The result is a list of tagged tokens.

from transformers import AutoTokenizer, AutoModelForTokenClassification
from transformers import pipeline

tokenizer = AutoTokenizer.from_pretrained("dslim/bert-base-NER")
model = AutoModelForTokenClassification.from_pretrained("dslim/bert-base-NER")

nlp = pipeline("ner", model=model, tokenizer=tokenizer)
example = "My name is Wolfgang and I live in Berlin"

ner_results = nlp(example)
print(ner_results)

Tool | What it does
AutoTokenizer | Splits the text into the tokens the model expects
AutoModelForTokenClassification | Loads the model that labels each token
pipeline with ner | Runs both steps and returns the tagged tokens

Example: Pass a sentence that names a person and the city where they live, as the sample above does. The pipeline returns a list of tagged tokens, where the name should come back as a person and the city as a location.

For newcomers and experts: If you are new, named entity recognition means finding the names of people, places and organizations in text. Run the sample, then replace the sentence with one of your own and read the tags.

For experts, the labels use a begin and continue scheme, so two entities of the same kind in a row can be told apart. The card reports test scores of 91.3 F1, 90.7 precision and 91.9 recall, a little below the original BERT results. Subword pieces are sometimes tagged as entities, so you may need to merge or filter the output.

Load the tokenizer and the model
Build an ner pipeline
Pass in a sentence of English text
Each token receives an entity label
Merge or filter the tags as needed

Title: Test scores from the card
F1: 91.3 points
Precision: 90.7 points
Recall: 91.9 points

Good to know: The points below cover limits and costs.

• Training data: English wire stories collected for CoNLL-2003, so text from other fields may be tagged less well.
• License: MIT.
• Cost to run: no API key or paid service is named.
• Data it reaches: it reads the text you pass in. Downloading the model needs a connection to Hugging Face.
• Not for: languages other than English, or texts that differ a lot from news. Subword tagging may need extra clean-up.

On Hugging Face, 13 November 2026

• Task: token classification
• Library: transformers
• Parameters: 108M
• Weights: about 433 MB in F32; about 54 MB at 4-bit (estimate)
• Formats: Safetensors
• Downloads in the last 30 days: 1,038,928
• Likes: 750
• License: mit
• Languages: en
• Quantized versions: 9
• Fine-tunes: 41
• Tags: transformers, pytorch, onnx, safetensors, bert, token-classification, model-index

Links

• Model: https://huggingface.co/dslim/bert-base-NER
• Files: https://huggingface.co/dslim/bert-base-NER/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download dslim/bert-base-NER

Then follow the usage in its model card. Nothing in this kit runs by itself.