keyphrase-mpnet-v1

This is a sentence-transformers model specialized for phrases: It maps phrases to a 768 dimensional dense vector space and can be used for tasks like clustering or semantic search. In the original paper, this model is used for calculating semantic-based evaluation metrics of keyphrase models.

This model is based on sentence-transformers/all-mpnet-base-v2 and further fine-tuned on 1 million keyphrase data with SimCSE.

Citing & Authors

Paper: KPEval: Towards Fine-grained Semantic-based Keyphrase Evaluation

@inproceedings{wu-etal-2024-kpeval,
    title = "{KPE}val: Towards Fine-Grained Semantic-Based Keyphrase Evaluation",
    author = "Wu, Di  and
      Yin, Da  and
      Chang, Kai-Wei",
    editor = "Ku, Lun-Wei  and
      Martins, Andre  and
      Srikumar, Vivek",
    booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
    month = aug,
    year = "2024",
    address = "Bangkok, Thailand and virtual meeting",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2024.findings-acl.117",
    pages = "1959--1981",
}

Usage (Sentence-Transformers)

Using this model becomes easy when you have sentence-transformers installed:

pip install -U sentence-transformers

Then you can use the model like this:

from sentence_transformers import SentenceTransformer
phrases = ["information retrieval", "text mining", "natural language processing"]

model = SentenceTransformer('uclanlp/keyphrase-mpnet-v1')
embeddings = model.encode(phrases)
print(embeddings)

Usage (HuggingFace Transformers)

Without sentence-transformers, you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings.

from transformers import AutoTokenizer, AutoModel
import torch


#Mean Pooling - Take attention mask into account for correct averaging
def mean_pooling(model_output, attention_mask):
    token_embeddings = model_output[0] #First element of model_output contains all token embeddings
    input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
    return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)


# Sentences we want sentence embeddings for
phrases = ["information retrieval", "text mining", "natural language processing"]

# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained('uclanlp/keyphrase-mpnet-v1')
model = AutoModel.from_pretrained('uclanlp/keyphrase-mpnet-v1')

# Tokenize sentences
encoded_input = tokenizer(phrases, padding=True, truncation=True, return_tensors='pt')

# Compute token embeddings
with torch.no_grad():
    model_output = model(**encoded_input)

# Perform pooling. In this case, mean pooling.
sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])

print("Phrase embeddings:")
print(sentence_embeddings)

Training

The model is trained on phrases from four keyphrase datasets covering a wide range of domains.

The model was trained with the parameters:

DataLoader:

torch.utils.data.dataloader.DataLoader of length 2025 with parameters:

{'batch_size': 512, 'sampler': 'torch.utils.data.sampler.RandomSampler', 'batch_sampler': 'torch.utils.data.sampler.BatchSampler'}

Loss:

sentence_transformers.losses.MultipleNegativesRankingLoss.MultipleNegativesRankingLoss with parameters:

{'scale': 20.0, 'similarity_fct': 'cos_sim'}

Parameters of the fit()-Method:

{
    "epochs": 1,
    "evaluation_steps": 0,
    "evaluator": "NoneType",
    "max_grad_norm": 1,
    "optimizer_class": "<class 'torch.optim.adamw.AdamW'>",
    "optimizer_params": {
        "lr": 1e-06
    },
    "scheduler": "WarmupLinear",
    "steps_per_epoch": null,
    "warmup_steps": 203,
    "weight_decay": 0.01
}

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 12, 'do_lower_case': False}) with Transformer model: MPNetModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})
)