camembertav2-base-name-classifier-v2

This model is a fine-tuned version of almanach/camembertav2-base on ele-sage/person-company-names-classification dataset.

It achieves the following results on the evaluation set:

• Loss: 0.0215
• Accuracy: 0.9942
• Precision: 0.9985
• Recall: 0.9910
• F1: 0.9947

Model description

This model is a high-performance binary text classifier, fine-tuned from camembertav2-base. Its purpose is to distinguish between a person's name and a company/organization name with high accuracy.

Direct Use

This model is intended to be used for text classification. Given a string, it will return a label indicating whether the string is a Person or a Company.

from transformers import pipeline

classifier = pipeline("text-classification", model="ele-sage/camembertav2-base-name-classifier-v2")

results = classifier([
    "Satya Nadella",
    "Global Innovations Inc.",
    "Martinez, Alonso"
])

for result in results:
    print(f"Text: '{result['text']}', Prediction: {result['label']}, Score: {result['score']:.4f}")

Downstream Use

This model is a key component of a two-stage name processing pipeline. It is designed to be used as a fast, efficient "gatekeeper" to first identify person names before passing them to a more complex parsing model, such as ele-sage/distilbert-base-uncased-name-splitter.

Out-of-Scope Use

• This model is not a general-purpose classifier. It is highly specialized for distinguishing persons from companies and will not perform well on other classification tasks (e.g., sentiment analysis).

Bias, Risks, and Limitations

• Geographic & Cultural Bias: The training data is heavily biased towards North American (Canadian) person names and Quebec-based company names. The model will be less accurate when classifying names from other cultural or geographic origins.
• Ambiguity: Certain names can legitimately be both a person's name and a company's name (e.g., "Ford"). In these cases, the model makes a statistical guess based on its training data, which may not always align with the specific context.
• Data Source: The person name data is derived from a Facebook data leak and contains noise. While a rigorous cleaning process was applied, the model may have learned from some spurious data.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 128
• eval_batch_size: 128
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
0.0354	0.0359	2000	0.0313	0.9915	0.9973	0.9874	0.9923
0.0388	0.0718	4000	0.0369	0.9884	0.9934	0.9858	0.9896
0.0254	0.1076	6000	0.0279	0.9927	0.9981	0.9888	0.9934
0.0296	0.1435	8000	0.0251	0.9930	0.9980	0.9895	0.9937
0.0241	0.1794	10000	0.0243	0.9932	0.9980	0.9898	0.9939
0.027	0.2153	12000	0.0252	0.9933	0.9972	0.9907	0.9940
0.0245	0.2511	14000	0.0252	0.9934	0.9986	0.9896	0.9941
0.0259	0.2870	16000	0.0244	0.9934	0.9972	0.9909	0.9941
0.0255	0.3229	18000	0.0245	0.9934	0.9974	0.9907	0.9940
0.0238	0.3588	20000	0.0244	0.9934	0.9968	0.9913	0.9941
0.0253	0.3946	22000	0.0240	0.9936	0.9977	0.9908	0.9943
0.0242	0.4305	24000	0.0240	0.9937	0.9976	0.9910	0.9943
0.0234	0.4664	26000	0.0225	0.9937	0.9987	0.9899	0.9943
0.0209	0.5023	28000	0.0235	0.9938	0.9987	0.9901	0.9944
0.0243	0.5382	30000	0.0226	0.9938	0.9975	0.9914	0.9944
0.0195	0.5740	32000	0.0235	0.9939	0.9981	0.9909	0.9945
0.0218	0.6099	34000	0.0225	0.9939	0.9979	0.9912	0.9946
0.0207	0.6458	36000	0.0229	0.9939	0.9985	0.9906	0.9945
0.0243	0.6817	38000	0.0219	0.9940	0.9988	0.9905	0.9946
0.0239	0.7175	40000	0.0218	0.9940	0.9987	0.9905	0.9946
0.0225	0.7534	42000	0.0216	0.9941	0.9983	0.9911	0.9947
0.0224	0.7893	44000	0.0218	0.9941	0.9984	0.9910	0.9947
0.02	0.8252	46000	0.0216	0.9941	0.9984	0.9910	0.9947
0.0268	0.8610	48000	0.0213	0.9941	0.9985	0.9910	0.9947
0.0254	0.8969	50000	0.0214	0.9942	0.9983	0.9912	0.9948
0.0229	0.9328	52000	0.0215	0.9941	0.9985	0.9910	0.9947
0.0224	0.9687	54000	0.0215	0.9942	0.9985	0.9910	0.9947

Framework versions

• Transformers 4.57.1
• Pytorch 2.9.0+cu128
• Datasets 4.4.1
• Tokenizers 0.22.1