paraphrase-filipino-mpnet-base-v2

paraphrase-filipino-mpnet-base-v2

This is a sentence-transformers model: It maps sentences & paragraphs to a 768 dimensional dense vector space and can be used for tasks like clustering or semantic search.

This model was trained using the student--teacher approach outlined in Reimers and Gurevych (2020).

The teacher model was sentence-transformers/paraphrase-mpnet-base-v2, and the student model was sentence-transformers/paraphrase-multilingual-mpnet-base-v2, which is based on XLM-R. We trained the model for 2 epoch using a batch size of 64 on parallel data English--Tagalog and English--Filipino data from OPUS. We found the data to be of variable quality and filtered it to only include sentence pairs that the Compact Language Detection kit (CLDv3) identified reliably as being in Tagalog or Filipino. Other parameters were left unchanged from the example make_multilingual_sys.py code in the sentence-transformers code base.

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
from scipy.spatial import distance
import itertools

model = SentenceTransformer('meedan/paraphrase-filipino-mpnet-base-v2')

sentences = ["saan pong mga lugar available ang pfizer vaccine? Thank you!","Ask ko lang po saan meron available na vaccine","Where is the vaccine available?"]
embeddings = model.encode(sentences)
dist=[distance.cosine(i,j) for i,j in itertools.combinations(embeddings,2)]
print(dist)

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
sentences = ['This is an example sentence', 'Each sentence is converted']

# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained('{MODEL_NAME}')
model = AutoModel.from_pretrained('{MODEL_NAME}')

# Tokenize sentences
encoded_input = tokenizer(sentences, 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("Sentence embeddings:")
print(sentence_embeddings)

Evaluation Results

We machine translated the STS data from SentEval to Filipino using the Google Translation API and used this for evaluation alongside the original English-language STS data. We used Spearman's rank correlation coefficient. We found roughly the same performance as the original base model (sentence-transformers/paraphrase-multilingual-mpnet-base-v2) on English while substantial gains were made for Filipino. For English, the average correlation is 0.80. For Filipino, it is 0.75.

For an automated evaluation of this model, see the Sentence Embeddings Benchmark: https://seb.sbert.net

Training

The model was trained with the parameters:

DataLoader:

torch.utils.data.dataloader.DataLoader of length 79097 with parameters:

{'batch_size': 64, 'sampler': 'torch.utils.data.sampler.RandomSampler', 'batch_sampler': 'torch.utils.data.sampler.BatchSampler'}

Loss:

sentence_transformers.losses.MSELoss.MSELoss

Parameters of the fit()-Method:

{
    "epochs": 2,
    "evaluation_steps": 0,
    "evaluator": "NoneType",
    "max_grad_norm": 1,
    "optimizer_class": "<class 'transformers.optimization.AdamW'>",
    "optimizer_params": {
        "correct_bias": false,
        "eps": 1e-06,
        "lr": 2e-05
    },
    "scheduler": "WarmupLinear",
    "steps_per_epoch": null,
    "warmup_steps": 10000,
    "weight_decay": 0.01
}

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: XLMRobertaModel 
  (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})
)

Citing & Authors

<!--- Describe where people can find more information -->