Graphlet-AI/eridu

Deep fuzzy matching people and company names for multilingual entity resolution using representation learning... that incorporates a deep understanding of people and company names and can work much better than string distance methods!

This is a sentence-transformers model finetuned from sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2 for person and company name matching using the Open Sanctions matcher training data. It maps sentences & paragraphs to a 384-dimensional dense vector space and can be used as part of a deep, fuzzy entity resolution process.

NOTE: this model is a work in progress. It is not yet ready for production use!

Model Details

TLDR: 5 Lines of Code

from sentence_transformers import SentenceTransformer


# Download from the 🤗 Hub
model = SentenceTransformer("Graphlet-AI/eridu")

names = [
    "Russell Jurney",
    "Russ Jurney",
    "Русс Джерни",
]

embeddings = model.encode(names)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

print(similarities.numpy())
# [[0.9999999  0.99406826 0.99406105]
#  [0.9940683  1.         0.9969202 ]
#  [0.99406105 0.9969202  1.        ]]

Project Eridu Overview

This project is a deep fuzzy matching system for person and company names for entity resolution using representation learning. It is designed to match people and company names across languages and character sets, using a pre-trained text embedding model from HuggingFace that we fine-tune using contrastive learning on 2 million labeled pairs of person and company names from the Open Sanctions Matcher training data. The project includes a command-line interface (CLI) utility for training the model and comparing pairs of names using cosine similarity.

Matching people and company names is an intractable problem using traditional parsing based methods: there is too much variation across cultures and jurisdictions to solve the problem by humans programming. This results in complex, cost prohibitive enterprise solutions for name matching like IBM InfoSphere Global Name Management. Machine learning is used on problems like this one of cultural relevance, where the time to manually programming a solution appproaches infinity, to automatically write a program. Since 2008 there has been an explosion of deep learning methods that automate feature engineering via representation learning methods including such as text embeddings.

This project loads the pre-trained paraphrase-multilingual-MiniLM-L12-v2 paraphrase model from HuggingFace and fine-tunes it for the name matching task using contrastive learning on more than 2 million labeled pairs of matching and non-matching (just as important) person and company names from the Open Sanctions Matcher training data to create a deep fuzzy matching system for entity resolution.

This model is available on HuggingFace Hub as Graphlet-AI/eridu and can be used in any Python project using the Sentence Transformers library in five lines of code. The model is designed to be used for entity resolution tasks, such as matching people and company names across different languages and character sets when matching records.

Model Description

• Model Type: Sentence Transformer
• Base model: sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
• Maximum Sequence Length: 128 tokens
• Output Dimensionality: 384 dimensions
• Similarity Function: Cosine Similarity
• Language: en
• License: apache-2.0

Model Sources

• Documentation: Graphlet-AI/eridu Documentation
• Repository: Graphlet-AI/eridu on GitHub
• Hugging Face: Graphlet-AI/eridu on Hugging Face
• PyPi Package: Graphlet-AI/eridu on PyPi

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("Graphlet-AI/eridu")

# Run inference
sentences = [
    'Schori i Lidingö',
    'Yordan Canev',
    'ကားပေါ့ အန်နာတိုလီ',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Binary Classification

• Dataset: sentence-transformers-paraphrase-multilingual-MiniLM-L12-v2
• Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.9885
cosine_accuracy_threshold	0.7183
cosine_f1	0.9825
cosine_f1_threshold	0.7086
cosine_precision	0.9782
cosine_recall	0.9868
cosine_ap	0.9971
cosine_mcc	0.9739

Training Details

Training Dataset

Unnamed Dataset

• Size: 2,130,621 training samples

• Columns: sentence1, sentence2, and label

• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	label
  type	string	string	float
  details	min: 3 tokens mean: 9.32 tokens max: 57 tokens	min: 3 tokens mean: 9.16 tokens max: 54 tokens	min: 0.0 mean: 0.34 max: 1.0

• Samples:

  sentence1	sentence2	label
  캐스린 설리번	Kathryn D. Sullivanová	1.0
  ଶିବରାଜ ଅଧାଲରାଓ ପାଟିଲ	Aleksander Lubocki	0.0
  Пырванов, Георги	アナトーリー・セルジュコフ	0.0

• Loss: ContrastiveLoss with these parameters:

  {
      "distance_metric": "SiameseDistanceMetric.COSINE_DISTANCE",
      "margin": 0.5,
      "size_average": true
  }
  

Evaluation Dataset

Unnamed Dataset

• Size: 2,663,276 evaluation samples

• Columns: sentence1, sentence2, and label

• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	label
  type	string	string	float
  details	min: 3 tokens mean: 9.34 tokens max: 102 tokens	min: 4 tokens mean: 9.11 tokens max: 100 tokens	min: 0.0 mean: 0.33 max: 1.0

• Samples:

  sentence1	sentence2	label
  Ева Херман	I Xuan Karlos	0.0
  Кличков Андрій Євгенович	Андрэй Яўгенавіч Клычкоў	1.0
  Кинах А.	Senator John Hickenlooper	0.0

• Loss: ContrastiveLoss with these parameters:

  {
      "distance_metric": "SiameseDistanceMetric.COSINE_DISTANCE",
      "margin": 0.5,
      "size_average": true
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: epoch
• per_device_train_batch_size: 768
• per_device_eval_batch_size: 768
• gradient_accumulation_steps: 4
• learning_rate: 3e-05
• weight_decay: 0.01
• num_train_epochs: 4
• warmup_ratio: 0.1
• fp16: True
• load_best_model_at_end: True
• optim: adafactor

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: epoch
• prediction_loss_only: True
• per_device_train_batch_size: 768
• per_device_eval_batch_size: 768
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 4
• eval_accumulation_steps: None
• torch_empty_cache_steps: None
• learning_rate: 3e-05
• weight_decay: 0.01
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 4
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.1
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• save_safetensors: True
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• no_cuda: False
• use_cpu: False
• use_mps_device: False
• seed: 42
• data_seed: None
• jit_mode_eval: False
• use_ipex: False
• bf16: False
• fp16: True
• fp16_opt_level: O1
• half_precision_backend: auto
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• local_rank: 0
• ddp_backend: None
• tpu_num_cores: None
• tpu_metrics_debug: False
• debug: []
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_prefetch_factor: None
• past_index: -1
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: True
• ignore_data_skip: False
• fsdp: []
• fsdp_min_num_params: 0
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• tp_size: 0
• fsdp_transformer_layer_cls_to_wrap: None
• accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adafactor
• optim_args: None
• adafactor: False
• group_by_length: False
• length_column_name: length
• ddp_find_unused_parameters: None
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• skip_memory_metrics: True
• use_legacy_prediction_loop: False
• push_to_hub: False
• resume_from_checkpoint: None
• hub_model_id: None
• hub_strategy: every_save
• hub_private_repo: None
• hub_always_push: False
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_inputs_for_metrics: False
• include_for_metrics: []
• eval_do_concat_batches: True
• fp16_backend: auto
• push_to_hub_model_id: None
• push_to_hub_organization: None
• mp_parameters:
• auto_find_batch_size: False
• full_determinism: False
• torchdynamo: None
• ray_scope: last
• ddp_timeout: 1800
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• include_tokens_per_second: False
• include_num_input_tokens_seen: False
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• eval_on_start: False
• use_liger_kernel: False
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	sentence-transformers-paraphrase-multilingual-MiniLM-L12-v2_cosine_ap
-1	-1	-	-	0.7140
0.7207	500	0.038	-	-
0.9989	693	-	0.0028	0.9911
1.4425	1000	0.0128	-	-
1.9989	1386	-	0.0021	0.9956
2.1643	1500	0.0084	-	-
2.8850	2000	0.0065	-	-
2.9989	2079	-	0.0015	0.9968
3.6068	2500	0.0056	-	-
3.9989	2772	-	0.0014	0.9971

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.12.9
• Sentence Transformers: 3.4.1
• Transformers: 4.51.3
• PyTorch: 2.7.0+cu126
• Accelerate: 1.6.0
• Datasets: 3.6.0
• Tokenizers: 0.21.1

Citation

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}

ContrastiveLoss

@inproceedings{hadsell2006dimensionality,
    author={Hadsell, R. and Chopra, S. and LeCun, Y.},
    booktitle={2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)},
    title={Dimensionality Reduction by Learning an Invariant Mapping},
    year={2006},
    volume={2},
    number={},
    pages={1735-1742},
    doi={10.1109/CVPR.2006.100}
}